University of Southern California Dissertations and Theses

Sunsetting: platform closure and the construction of digital cultural loss

(USC Thesis Other)

Sunsetting: platform closure and the construction of digital cultural loss

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

SUNSETTING:

PLATFORM CLOSURE AND THE CONSTRUCTION OF DIGITAL CULTURAL LOSS

by

Frances R. Corry

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(COMMUNICATION)

August 2022

Copyright 2022 Frances R. Corry
ii
Table of Contents

Acknowledgments………………………………………………………………………………………...iv

List of Tables……………………………………………………………………………..…….….……..vii

List of Figures………………………………………………………………………………..…………..viii

Abstract…………………..………………………………………………………………………………..ix

Introduction………………………………………………………………………………………………...1
A Digital Dark Age? …………………………………………………………..……………………4
Platforms Addressed in this Project………………………………………………………………....7
Key Contributions………………………...………………………………………………….…….16
Key Terms: Platform, Content, and Sunsetting………………………………………………….…22
Dissertation Structure……………………………………………………………………………...27

Chapter 1: Memory, Media, Man, Machines: Becoming Digital Memory Infrastructure………...….29
Theories of Memory and Media in the 20th Century……………………………………….………32
Machines and Memory at Mid-Century…………………………………………………….………40
Ubiquitous Computing in the Early Aughts…………………………………………………….….48
Platforms Embrace Memory………………………………………………………………….……54
Digital Memory Infrastructure……………………………………………………………….….…60

Chapter 2: Vicissitudes of Value: The Commercial Logics Behind Digital Loss………………..……63
Rubbish Theory: On Salvaging Value………………………………………………………….…...67
The Breakdown and Decay of Infrastructure……………………………………………………....72
Friendster: For Whom Not all Hockey Sticks are Equal………………………………………...….75
MySpace: The Rise, Fall, and Attempted Resuscitation of Value ……………………………..……86
How Valuation Structures Commercial Digital Loss……………………………………….………99

Chapter 3: Social Values and the Shaping of Platform Afterlives……………………………………..102
From Value to Values in Sunsetting……………………………………………….……………...103
What’s at Stake When We Talk About Platforms and Values……………………….…………….107
Seeing Values………………………………………………………………………….………….110
MySpace: The Thrill of Creative Destruction …………………………………………………….111
Vine: Choosing the ‘Right Way’ to Shut Down a Platform ……………………………….………127
Sunsetting as a Value-Laden Act….………………………………………………………………140

Chapter 4: The Construction of Destruction: Material Considerations in Sunsetting……….………144
The Construction of Destruction: Between Materiality and Constructivism …………………..….145
Dismantling Software Systems……………………………………………………………………147
MySpace: Technical Debt to Total Destruction……………………………………………...……150
Vine: Pulling Apart a Platform, Assembling an Archive……………………………………..……167

Chapter 5: Sunsetting, a Story: Media Accounts of Platform Closure ……………………………..…176
Biographies of Breakage …………………………...………………………………….………….179
Infrastructural Expectations ………………………….………………………….……………….181
Narrating Sunsetting…………………………………….………………………………..………183
Permanence, Ephemerality, and Expectations of Digital Memory Infrastructure…………………191
For Whom Does Sunsetting Matter? ………………………………………………..……………196
iii

Conclusion ………………………………………………………………………………………………199
Toward Deletion Studies……………………….……………………………..………..…………202
Alternatives: Stewardship and Speculation……………………….…………....………..…………205

References ……………………………………………………………………….……………………...210

Appendix A: Methods…………………………………………………………………...………………242
Choice of Platforms………………………………………………………………………………242
Document Analysis…………………………………………………………………...…………..244
Interviews and Thematic Analysis………………………………….………………………..……245
Effect of COVID-19 Crisis on Methods……………………………….…………………………255

Appendix B: Interview Questions………………………………………………………………………260

iv
Acknowledgements

Of any part of the dissertation, I’ve looked forward to writing this section the most. I am not
exaggerating when I say I could write pages about each person I’ll mention here; what follows is
necessarily incomplete, not only because there are far fewer pages than there could be but inevitably
because I missed someone who matters. So: please insert your name here if you don’t see it in what
follows: ________________________.
This project is dedicated–– and owes much to––my two grandmothers, Corinne Endreny
Kirchner and the late Patricia Murphy Corry, both educators and sources of intellectual inspiration.
Patricia Corry showed me the pleasures of words, the value of kindness, and the delights of beating
other people at cards. My ‘gma’ Corinne, thank you for paving the way for this degree, for taking me
in all those summers and being my roommate, for the always stimulating conversation, and for your
good humor and love.
I couldn’t have asked for a better advisor than Christina Dunbar-Hester. As a first year, I
had hoped Christina would work with me because I found her research, her writing, and her
feedback so astute, and suspected she could push my own work in ways I would value. (I was right
about this!) But I also got an advisor who I trust deeply, and whose confidence in me helped inspire
that in myself.
My committee members Mike Ananny and Geof Bowker were outstanding guides for this
project and much beyond. I’ve benefitted from the Mike’s mentorship and generosity since my early
years in this program, and I’m grateful to him for his considered intellectual guidance and unfailing
kindness. Geof Bowker’s work has shaped my own before we even had the chance to meet. I am
better as a scholar and colleague because of his expansive ways of thinking and overall care.
There are many other faculty at USC whose encouragement and expertise were instrumental
to this project and to my graduate school experience. These folks include Josh Kun, who I was lucky
v
enough to RA for for many years; Peter Monge, my first advisor; Jennifer Petersen, who served on
my qualifying exam committee; Larry Gross, who guided me at IJoC; and Kate Crawford, who
helped me expand my research in my last year at USC. Administrative staff at Annenberg, especially
Anne Marie Campian and Sarah Holterman, have provided unwavering support over the last six
years, and my dissertation and graduation (along with everyone’s in this program) are testaments to
their skill and dedication.
I’ve been fortunate to have mentors from what people say is a USC rival but who seemed to
have put it aside to offer me guidance over the years: Safiya Umoja Noble and Sarah Roberts of
UCLA. I’ve also never stopped being grateful for the faculty at Barnard College who first
encouraged my writing and research: especially Jonathan Beller, Yvette Christiansë, and Richard
Panek.
My colleagues that became close friends at USC: I am so lucky to have gone through this
program with you. My cohort, the ‘PhDivas,’ opened my eyes to the breadth of communication
scholarship and most importantly to what academic community and support can look like.
Folks in cohorts across the program were sources of collaboration and community, and I’m
deeply thankful to them for their guidance and friendship. This included the ‘Terrazza squad:’ Matt
Bui, Courtney Cox, Caitlin Dobson, Brooklyne Gipson, Sulafa Zidani, and Simogne Hudson. It also
included the ‘tree growing’ group–especially my former neighbors Rachel Moran and Sulafa (again!),
and Lauren Sowa–who were constant sources of support and fun. I’m also grateful for the
friendship of folks like Sierra Bray (and Pete!), MC Forelle, Olivia Gonzalez, Jessica Hatrick, Ana
Howe Bukowski, Ed Kang, Paulina Lanz, Kate Miltner, Clare O’Connor, Cerianne Robertson, and
Hamsini Sridharan. My accountability group–Sophia Baik, Donna Kim, Andrea Alarcon and Sole
Altrudi–our Monday mornings, and your friendship and guidance, have been very (very!) bright
vi
spots over the last few years. A final note to my ‘pandas’ and academic sisters Sole and Andrea: what
did I do to deserve you?!
My friends from far and wide outside of graduate school have grounded me in so many
ways, especially the now more than decade long bonds with my soul/hallmates from Barnard who
are now in LA: Ashton Cooper, Margaret Boykin, and Lori Goldman. Judith G deserves special
mention for her care and for teaching me so much over the years.
Somewhere along the family and friend spectrum: Beth Corry and Reggie Guerrero, you
made LA feel like home, hence why Alex and I moved closer and closer to you every year. Our
greatest sadness in leaving here is leaving the two of you.
My family, who has been there for me through the highs but especially the lows, thanks. I’m
lucky to have you and to have been in lifelong company with people so caring and funny and with
such thick hair: Mom, Dad, Maggie, Grace, Aaron, Riley, Hannah, and sometime soon that other
baby. I’m also lucky to have the support of the Konrad family: Rudy, Karin, Annika, Andy, baby
Otto, Kristine, Brooks, Wynn and Clara.
Finally, to Alex Konrad: for you I must be most minimal at risk of being most maximal.
Thank you for everything.

vii
List of Tables

Table 1: Participants Per Platform …………………………………………………………………8

Table 2: Collected Media Accounts of Platform Closure…………………………………………178

Table 3: Interviewees……………………………………………………………………….……257

viii
List of Figures

Figure 1: Friendster timeline………………………………………………………………………13

Figure 2: MySpace timeline………………………………………………………………………..14

Figure 3: Vine timeline…………………………………………………………………….………15

Figure 4: Hockey stick graph………………………………………..……………………………..65

Figure 5: Wooden bench in MySpace lobby…..………………………..…………………..…..…119

Figure 6: Creating and destroying a sand mandala………………………………………………..145

Figure 7: MySpace's FAQ for lost content……………………………………………………….159

Figure 8: Eight stills from the last Vine uploaded…………………………………………….…..173

ix
Abstract

Despite dominant cultural narratives about platform vitality, whether networks’ global
penetration or companies’ overwhelming political and economic power, the history of social media
is marked by shutdown, failure and data loss. This has included sites from the mid- 1990s, like
GeoCities; networks popular in the early to mid-2000s, like Friendster and MySpace; and more
recent platforms like Vine and Google+. Other companies have shorn off user content as their
platforms have changed, due to either economic or legal pressure. These deletions often take place
with little warning to users and archivists, resulting in the loss of innumerable personal records–from
photos to music tracks, conversations to contacts.
This form of information loss differs from other media. Traditional paper-based media, for
instance, are often “preserved by neglect” (Rosenzweig, 2003)– a diary left in an attic may become
faded but can still be read years later. The public web, or websites that do not require password
protected accounts, can be saved through preservation software like that run by the Internet
Archive. Platform preservation or deletion, however, relies on the managerial and technical decisions
made by platform companies. Understood at scale, the ways that platforms shut down and handle
user records in this process will have significant consequences for both individual and shared
recollections of the past (see Couldry et al, 2014; De Kosnik, 2016).
The process of social media closure is colloquially called sunsetting, a term used to describe
the ending of a digital product, and it is this process that is examined in this dissertation. Through
semi-structured, in-depth interviews with over fifty former employees of platforms that have sunset,
including MySpace, Vine, GeoCities, and Friendster, along with the analysis of recent media and
historical documents, this project discusses the process and consequences of platform sunsetting as
it has been practiced to date.
x
Drawing together studies of media, information technology, and memory, Sunsetting observes
how organizations and employees make decisions about what to do with no-longer useful data and
discusses how the fate of this user-generated content shapes memory practices into the future. In
doing so, it first offers a descriptive contribution by providing a behind-the-scenes look at production-
level platform cultures beyond much-scrutinized moments of innovation. It also contextualizes
sunsetting and frames its importance through the concept of digital memory infrastructure, arguing that
social media platforms invisibly support modern memory practices, while noting that people bring
infrastructural expectations, or normative beliefs about how an infrastructure functions, to their
interactions with it. Finally, and most crucially, it proposes commercial digital loss, showing how
platform cultures’ extractive logics come to structure content loss, in turn shaping the construction
of the mediated past. The dissertation ultimately argues that the process of sunsetting is a powerful
but less scrutinized mechanism through which platform organizations shape public discourse.
Through the construction of digital loss, the mediated past is produced according to forms of value
and valuation endemic to the US techno-cultures that are examined in this project.

1
Introduction
On May 9 of 2019, it was announced that MySpace–the formerly dominant social
networking site–had lost 12 years of user data in what the company chalked up to a “botched server
migration.” What was lost included some 53 million songs produced by 14 million artists, along with
unreported numbers of user profiles uploaded between 2003 and 2015. Some referred to this
deletion as lost history, the massive erasure of “modern folklore” (Scott, quoted in Chokshi, 2019).
Musicians who had used the site as a storage mechanism for their work bemoaned the sudden loss
of their archive. Some users grieved the ability to look back, the occasional glimpse at a past self.
Others would implicate the company: Andy Baio, a technology journalist and entrepreneur who first
reported the loss, was skeptical that it was an accident. “Flagrant incompetence may be bad PR, but
it still sounds better than ‘we can’t be bothered with the effort and cost of migrating and hosting 50
million old MP3s.’” And still, some seemed to shrug. These were the dynamics of digitality, it
seemed, and we should get used to their corporate vagaries and patterns of loss (Krukowski, 2019).
While MySpace’s data loss was remarkable for its vast scale–having at one point been the
largest repository for music online, and the most visited site in the United States–the partial or
wholesale shuttering of social media platforms is not unusual. Despite dominant cultural narratives
about social media platform vitality, whether immense global diffusion or companies’ overwhelming
political and economic power, their history is marked by shutdown and failure. Companies and
platforms shutter with understated regularity. This has included sites from the mid-1990s, like
GeoCities, which hosted millions of user-created pages. It has included networks popular in the
early 2000s, like Friendster, which relied on the sharing of hobbies, photos, and friend networks, and
centered around dating. It has included Vine, the video-sharing platform owned by Twitter, which
had been a primary site for cultural production by young Black content creators, and Google+,
which was the search engine’s response to networks like Twitter and Facebook. Other platforms

2
have shorn off user content as their platforms have changed, due to either economic or legal
pressure. Flickr, the photo-sharing site, for instance, deleted photos from free user accounts that
contained more than one thousand images in 2019. Tumblr, a blogging platform and important site
for alternative erotic communities, deleted all blogs containing adult content in December of 2018.
And these are just platforms based out of the United States: South Korea’s CyWorld, Sweden’s
LunarStorm, China’s Jiepang, and others have suffered similar fates.
MySpace’s 2019 content loss occurred when I was beginning to focus on platform closure as
a research subject in earnest. The fate of user-generated content is what had piqued my interest in
this topic in the first place. As someone who had previously used traditional archives housed in
libraries and museums for historical research, I wondered how the diverse media generated and
hosted online today would be organized, maintained, and used by researchers in the future. The
media that users uploaded to these sites, which were then stored and organized by a platform
company in seeming perpetuity, appeared similar to informal documentary traces–from scrapbooks
to diaries, home videos to theater tickets, ledgers and letters–that non-elites had left behind in
previous eras. Whether mobilized by trained historians, artists, archivists, or individuals, traces like
these have done much to shape an understanding of the evolution of power, social struggle, and
identity–especially in relation to race, gender, class, and sexuality–over time. How would the ongoing
present be made sense of in the future if media from platforms–which offers a window into how
people are grappling with historical events, engaging in political discourse, or participating in the
unfolding of regular life–were inaccessible? Moreover, memory scholars have shown how social
media platforms are used by communities to organize, reflect on, and shape the meaning of the past
(e.g. De Kosnik, 2016; Richardson & Hessey, 2009; Smit et al, 2017; Villa Nicholas, 2019). How
would data loss shape communities’–and the individuals that compose them–ability to reflect on
themselves?

3
In a single event, MySpace’s data loss seemed to illustrate the complexity of this issue
beyond the initial questions that had led to my interest in platform closure. It showed how data loss
on platforms was entangled in a web of social media organizations, platform users, user
expectations, technological economies, and physical technologies. It also revealed the ambiguity
around what it even meant for a platform to be ‘shut down,’ to have died. After all, while you could
still access the MySpace URL after this loss, the site was essentially unusable as a platform––and had
been for some time. Most of all, MySpace’s data loss sharply juxtaposed the utopic notions of user
agency that accompanied the debut of social media platforms––in which online platforms had
appeared to democratize expression, with individuals and communities able to supersede the
monolithic control of traditional media institutions (van Dijck, 2009)––and the apparent lack of
agency that platform users in fact possessed in deciding the long-term fate of this user-generated
content.
MySpace’s data loss had made the complexity of this issue apparent. What continued to be
unclear, however, was just how content loss and platform closure came to be. What happens at the
organizational level leading up to data deletion? What decisions are platform employees faced with
as platforms shut down or go into decline? What values and worldviews do they bring to this
process? How does the technical makeup of a platform influence these dynamics? That is, how are
losses like these constructed by platform organizations, platform employees, and platform
infrastructure over time? Answering these questions, it seemed, would allow both analysts and
broader publics to understand how social media platforms continue to be entangled in social life and
in public discourse even after they fall out of the public eye, and why these entanglements matter,
especially for historical and social reflection. In doing so, they would provide a basis for discussing
what the long-term fate of digital cultural production hosted on platforms should look like, offering
an opening for considering alternatives beside what seemed to be the default mode of deletion. By

4
focusing on ‘sunsetting,’ a colloquial term for the closure of a site or service, and data loss, this
dissertation attempts to answer these questions, to untangle the complexity of this issue, and to open
up space for considering how a platform’s end-of-life could be different.
A Digital Dark Age?
I was not the first to ask how changes in mediation might affect individual, community, and
historical reflection. My interest in platform closure and data loss grew from conversations
historians and archivists have been having since the advent of the popular web in the mid-1990s.
Since then, these professions have sounded the alarm that the online shift in the creation and
management of information could pose problems for both researchers and the public who may want
to retrieve and reflect on various records now and in the future. The late historian Roy Rosenzweig
(2003) notes that paper-based records are often “preserved by neglect:” a diary left in an attic may
become brittle or faded but can still be read years later. However, corollary personal records made
online have different patterns of loss: blog posts published on a personal website, for instance,
disappear if the domain expires. Turnover and expiration in the online world is indeed a common
occurrence, with the average age of a webpage just 100 days (Kneese, 2019).
The threat of information scarcity based on shifting modes of decay were phrased in salient
but dramatic terms during that era: as the threat of a “digital dark age” (Kuny, 1999). Since then,
however, relatively robust methods have been developed for archiving the public web, or sites that
are not locked behind password protected accounts. The Internet Archive’s Wayback Machine is the
most well-known web archiving application, while their service Archive-It is used by numerous
institutions, especially universities, to preserve records of their sites over time. There are many other
web archives: from the Library of Congress’ archiving of US government websites, to other national
archives like the UK Web Archive or the Icelandic Web Archive; university-run archival sites like
Stanford Archives or Harvard University’s perma.cc; and third party not-for-profits like Bibliothecha

5
Alexandrina and the End of Term Web Archive. While these sites are focused on archiving the
public web, platform scholars Anne Helmond and Fernando van der Vlist (2019) have argued that
web archives also provide overlooked resources for understanding platforms’ commercial and
technical evolution over time through public-facing information that can be captured through web
crawling. Accessing dynamic and ever-changing end-user content through these web-archiving
repositories, as they note, is another story. Platforms present different challenges and thornier
questions about digital preservation: they are typically locked behind private accounts, have technical
configurations that are difficult to save, or are held by companies that remain skittish about having
their content preserved by another entity. What happens to this user-generated content at-scale is
then largely left to decisions made by platform companies, not outside archivists.
That platform organizations have disproportionate sway in determining the fate of user-
generated content reflects what critical data scholars have observed as “extremely asymmetrical
power relations” in data management broadly speaking (D’Ignazio and Klein, 2020, p. 36). That is,
institutions, whether governmental or corporate or academic, who have the agency to collect data
about others hold more power in determining how this data is wielded than those populations about
whom data are routinely collected. Importantly, where I was locating these asymmetrical power
dynamics were not at stages when social media platforms’ practices–data or otherwise–are usually
under scrutiny, that is, when they are growing and flourishing and functioning well. Instead, I was
seeing these power dynamics at less-visible points in the ‘lifecycle’ of these sociotechnical systems:
asymmetrical power relations even when these platforms seemed to be declining, even when they
were closing, and even after they were presumed to be dead. Media scholar Abigail de Kosnik has
eloquently summarized this power differential when, writing about the closure of a platform she had
participated on, she asked (Clark et al, 2014, p. 1455):

6
What will happen when Tumblr folds? When Google sells off or closes YouTube? When
Instagram, Twitter, and Facebook go dark? The fate of all of the text, image, and video ever
posted to any of these platforms will not ultimately be decided by the millions upon millions
of users who made that content but by the corporations that own the platforms.

Observations like this cast light on the need to understand not just how content may be ‘consumed’
by historians, users, and archivists, but how these dynamics of loss are actively constructed at the
platform level. In turn, to understand platform closure and data loss, I directed my attention to
platform corporations and platform employees, to how they made decisions about the fate of user-
generated content during decline and closure, how these decisions shape what remains of these sites
going forward, and what the consequences of these decisions are.
This dissertation reflects sustained engagement with platform sunsetting through qualitative
interpretive methods, including the analysis of 56 interviews as well as the study of sunsetting-related
press and grey literature. Of these interviews, 53 were with former employees of platforms that have
shuttered. (The remaining three were conducted with people who work to archive or export
platform data.) I conducted these interviews in the aftermath of MySpace’s 2019 data loss, beginning
interviews in March of 2020 and continuing through January of 2021. Over these 11 months, I
spoke to employees from a range of defunct US-based social media platforms, including MySpace,
yes, but also GeoCities, Friendster, Vine, as well as smaller and less well-known platforms like Bolt,
Couple, MakeOutClub and 43Things. In this project, these interviews are put into conversation with
intersecting literatures from media studies, science and technology studies, and memory studies.
Together, they ultimately help inform the central argument of this project: that platform sunsetting
is an act of sociotechnical destruction which leads to a particular construction of the mediated past, one
shaped by the priorities and worldviews of the US tech industry.
In what follows in this Introduction, I give a brief synopsis of the platforms I focus on in
this project, chart the major contributions to scholarly literatures on platforms, digital memory, and

7
infrastructural breakdown, and then briefly address important terms that the project relies on,
including ‘content,’ ‘platform,’ and ‘sunsetting.’ To conclude, I discuss the structure of this
dissertation and preview the chapters that compose the remainder of it.
Platforms Addressed in this Project
This dissertation draws on interviews with employees from 10 different platforms (see Table
1) but focuses most significantly on three major platforms: Friendster, MySpace, and Vine. These
platforms were chosen for reasons I address in detail in Appendix A, choices that evolved as
interviews progressed during 2020 and 2021. In brief, I’ve focused on Friendster, MySpace, and
Vine, because they were all culturally significant platforms in the United States, with millions of
users and sustained user engagement. All three were identified as “social networks” or as “social
media platforms” when they existed. While similar in general cultural impact, these sites also offer
variety: they hail from slightly different eras and have slightly different foci. Friendster, launched in
2002 and at its peak popularity in the US in 2002-2003, is commonly thought to have been the
predecessor of MySpace. MySpace launched in 2003, enjoying its peak popularity between 2005-
2007, and was in turn thought of as the predecessor to Facebook. In the US, no profile-based social
network has gained as much market dominance since then. Vine, the mobile-first video platform
owned by Twitter that launched in 2013, offers an example of a platform that debuted as
smartphones became popular ways to access internet-based resources. Finally, and most
fundamentally, they are all sites that are defunct. In conversation with one another, they offer the
opportunity to compare processes of closure and data deletion, providing a foundation to
understand sunsetting not as an isolated occurrence but as a socio-technical phenomenon.
While the chapters of the dissertation provide background on each platform, I offer a brief
synopsis of Friendster, MySpace, and Vine here as well to orient the reader. These descriptions are
accompanied by a timeline for each platform (Figures 1, 2, and 3) chronicling major events.

8

Table 1. Participants Per Platform
Platform Number of participants
MySpace 18
Friendster 13
GeoCities 6
Vine 8
Bolt 2
Couple 1
Friends Reunited 1
Makeout Club 1
Foodily 1
43 Things 1

Friendster
Friendster, based in San Francisco, California, was founded in 2002 by entrepreneur
Jonathan Abrams. Early on, the site experienced extraordinary growth for that era, with three
million users by early 2003 (Ulunma, 2020). First built for those seeking relationships, Friendster
operated as both a dating site and platonic social network where users could connect with one
another while viewing their degrees of separation from other users in the social graph. (Friendster
helped popularize the term ‘social network’ because of this network-science based feature.) Much
like Facebook today, Friendster users were encouraged to use their real names and images on their
personal profiles, and could send messages, post photos, and eventually see others’ activity within

9
the network via a centralized newsfeed. Given Friendster’s sizable and affluent audience in the US,
investors from blue-chip Silicon Valley-based venture capital firms like Kleiner Perkins Caufield &
Byers, BenchMark Capital, and Battery Ventures viewed the site as having limitless potential. These
groups rapidly invested in Friendster, and then joined the board to help steer the company
(Arrington, 2006). In 2003, Google attempted to buy the site, but Friendster founder Abrams
rebuffed the offer.
Even in these early and successful years, however, Friendster experienced challenges. The
site had trouble handling the crush of user traffic and was frequently inaccessible because of this.
The company’s investor-led board soon pushed original founder Abrams out, and executives rotated
in and out of the CEO position. In the mid-2000s, US-based use tapered off as Friendster users
switched to MySpace and eventually to Facebook. The company also underwent numerous strategic
changes while attempting to live up to its early expectations.
While built around users in the United States, Friendster had also become popular in
Southeast Asia, including the Philippines, Malaysia, Indonesia, and Singapore, where its user base
rapidly exceeded that in the United States within the site’s earliest years. Based on interviews, the
company tried to ignore its expanding userbase in Southeast Asia, who appeared ‘unmonetizable’ to
Friendster’s investors. By 2006, Friendster was considered an also-ran by most US-based social
network users and technologists, but still experienced significant use in Southeast Asia. Rather than
shutting it down, which investors considered doing, they instead attempted to be acquired. The
platform would be sold to Malaysian payments company Money Online (MOL) in 2011 and turned
into a gaming site. In this transition, all user content was removed from the platform. The platform
was rendered defunct in 2015, suspending its services, and ultimately was dissolved as a company in
2018. I address Friendster’s story most concertedly in Chapter 2.

10
MySpace
MySpace followed in Friendster’s footsteps, its Los Angeles, California-based founders
inspired by the social network concept but wanting to create one that was dedicated to music,
allowing bands to list their gigs and post their songs. While it always retained a focus on music, the
site soon grew beyond that scope. By 2006, it was the most visited site in the United States
(Arrington, 2006). In contrast to Friendster, MySpace was known as a site where users could
customize their pages with colorful backgrounds, music, a display of one’s ‘Top 8’ friends, and other
widgets, whether quizzes or blogs. Aside from this customization, users could post photos in
albums, connect with other users by becoming ‘friends,’ send messages to others in the network, and
write comments on others’ profiles. The site, along with parent company Intermix Media, was
purchased by Rupert Murdoch’s News Corp in 2005 for $590 million dollars.
By 2008, however, MySpace’s popularity was showing signs of waning. In 2006, competitor
Facebook had opened to those outside the colleges and corporations it first served, and users were
starting to switch to that platform, sometimes using it exclusively. Users and the press talked about
Facebook’s comparatively ‘clean’ interface, bemoaning the custom aesthetics–and the numerous
ads–that MySpace pages had become known for. By late 2010, MySpace was a diminished version of
its former self, with about 55 million unique monthly users compared to Facebook’s 155 million
(Roy, 2011). The company attempted a re-design with a new logo and user interface, one that was
less customizable by users and more ‘clean,’ mimicking Facebook’s look. In 2011, they were
purchased by online advertising company Specific Media, who attempted to relaunch the site as a
portal for bands and other musical artists. Debuting in 2013, the site was a modest success before
losing steam, fizzling over the course of a year. Ultimately, the site was not deleted but was
increasingly untended, with no new original content or technical updates. The MySpace site stayed

11
online since then, but has changed hands numerous times, from Specific Media to Viant to Time
Inc. to Meredith Corp.
Since its creation, the MySpace name has been one haunted by data loss, whether intentional
or accidental. Before the URL was acquired by the company that launched MySpace
(eUniverse/Intermix), it was a cloud storage platform operated by a completely different company, a
site where users could set up an account to store their files online. (An article chronicling their
demise called this service an “online closet” [Morris, 2001]). In a sign of things to come, when
MySpace the storage company went under in 2001, it deleted “untold numbers of files stashed there
free by its registered users,” having given its users six days’ notice to retrieve their files (ibid). The
defunct cloud storage company estimated that they had deleted 7.5 million files as they sunset. In
2013, as MySpace re-launched under Specific Media, the company would sunset the ‘original’
MySpace, creating a new site where only user accounts and user photos were taken over from the
earlier version. Blogs, comments, widgets, and messages were deleted from existing user profiles.
Finally, in 2019, the user content that remained from the 2013 version also disappeared in the
‘botched server migration.’ I address MySpace in all chapters that draw on interviews, including
Chapters 2, 3, and 4.
Vine
Vine, the platform owned by New York City-based Twitter, launched auspiciously on
January 24, 2013, four years before it would fold on January 17, 2017. One of the first apps that
embraced mobile video, users would share six-second, looping videos with their followers, a format
enabled by the app’s easy-to-use editing tools. By October 2013, Vine had 40 million users, making
it the fastest-growing app in the world (Shontell, 2013). It became known as a place for rapid and
easily memed forms of entertainment, particularly a brand of frenetic comedy (Roettgers, 2016),
while the app’s videos gained cultural and often linguistic ubiquity (Romano, 2016). Vine users

12
became celebrities with agents and entertainment deals, several of them even living and creating
together at an apartment building in Los Angeles on the corners of Hollywood Boulevard and Vine
Street (Friend, 2014).
Vine’s golden era started to fade within a year. Competing platform Instagram debuted video
sharing in 2013, the press began to speculate about Vine’s waning interest from advertisers, and
rumors swirled of Twitter’s diminished investment in the app as the years wore on. They were
ultimately true. Vine announced its closing in a Medium blog post in October of 2016, its founders
noting that they would shut down the social network “the right way.” Diverse reasons were
speculated for why Vine was ultimately closed, but the consensus gleaned from interviews for this
project is that Twitter was attempting to become profitable, and any components that were not
obvious profit centers were at risk of being cut.
While all Vine employees were laid off as the sunset was announced, ten employees were
selected to stay on for three months to help close the app, ultimately building a web-based version
of Vine where users could access previously posted content. On the official date of its closing, Vine
debuted a camera app in place of its social network app and transformed its website (vine.co) into an
archive, where users could browse top Vine content by year and log-in to see their own Vines. On
May 25, 2018, the date of required compliance with the European Union’s General Data Protection
Regulation (GDPR), Vine removed browsable access to the site and the year-by-year compilations
they once had stored there, though individual files could still be accessed if one had the original links
that connected to the archive. Today, Vine.co greets visitors with a non-interactive landing page
displaying the Vine logo and an emoji, waving goodbye. Vine’s story is addressed in this project in
Chapters 3 and 4.

13

Figure 1. Friendster Timeline.

14

Figure 2. MySpace Timeline.

15

Figure 3. Vine Timeline.

16
Key Contributions
The stories of Friendster, MySpace, and Vine are woven together in this project, with
supporting evidence provided by the other seven platforms whose employees I spoke to. As I’ll
show here, their stories in concert speak to concerns at the intersection of media studies,
information studies, and science and technology studies. In speaking to these concerns, they make
three central contributions. The dissertation first offers a descriptive contribution by providing a
behind-the-scenes look of production-level platform cultures beyond much-scrutinized moments of
innovation. It also contextualizes sunsetting and frames its importance through the concept of digital
memory infrastructure, arguing that social media platforms invisibly support modern memory practices,
while noting that people bring infrastructural expectations, or normative beliefs about how an
infrastructure functions, to their interactions with it. Finally, it proposes commercial digital loss, showing
how platform cultures’ extractive logics come to structure content loss, in turn shaping the
construction of the mediated past. I explain these main contributions in greater detail here.
Beyond Innovation, Dispelling Platforms’ ‘Phantom Objectivity’
Looking to sunsetting contributes to a larger, ongoing story about platforms–common
shorthand for those online socio-technical assemblages that host user generated content (Gillespie,
2010)–and their evolving relationship to society. Over the last decade, studies that take critical
approaches to platforms have done much to show how extant platforms have ‘politics,’ meaning
that they embody and perpetuate specific forms of power (Winner, 1980). These studies have shown
the ways that extant platforms shape public values according to corporate interests (van Dijck, Poell
& de Waal, 2018), transform the nature and perception of public speech (Gillespie, 2017), perpetuate
existing systems of racial and gendered bias through algorithmic selection (Noble, 2018), and
beyond.

17
This project, while concerned with similar entanglements between platforms and society,
takes a slightly different tack, adopting what science and technology studies scholar Steven Jackson
(2014) has called “broken world thinking” to examine this relationship. This lens proposes that
“breakdown, dissolution, and change, rather than innovation, development, or design as
conventionally practiced and thought about are the key themes and problems facing new media and
technology scholarship today” (p. 222), an analytical move that tends to usurp ideologies of
innovation and progress mapped to new technology (Russell & Vinsel, 2018; Vinsel & Russell,
2020), reveal the persistent politics of that which has appeared to fade away, and shed light on
otherwise obfuscated material, social, and economic relations. In adopting this framing–which has
largely been used to examine artifacts like ships (Jackson, 2014), cell phones (Ahmed et al, 2015;
Houston et al, 2016), and ‘traditional’ infrastructure like roads and bridges (Anand et al, 2018)–and
bringing it to platforms, this project shows the analytical utility in applying perspectives from the
study of breakdown, maintenance, and repair to the study of communications platforms. It argues
that platforms–and indeed other sociotechnical systems–have politics not only during their
construction and active use, but indeed have politics even after these systems are widely understood as
dead or in disrepair. This project therefore contributes especially to critical studies of platforms by
providing an in-depth description of diverse platforms’ breakdown and dismantling, offering a more
comprehensive understanding of platforms’ sociotechnical entanglements beyond moments of
innovation.
Like other critical studies of communication platforms, this work strives to undo what
technology scholar Leo Marx, referencing philosopher George Lukacz, has called technology’s “aura
of phantom objectivity” (2010/1997, p. 576). ‘Phantom objectivity,’ Marx explains, is a quality of
technology that makes it appear as an autonomous thing–for instance, that transportation technology

18
is composed of roads and cars, or that information technology means computers and wires–rather
than an entity shaped first and foremost by social relations. He notes that:
By consigning technologies to the realm of things, this well-established iconography distracts
attention from the human—socioeconomic and political—relations which largely determine
who uses them and for what purposes. Because most technologies in our corporate capitalist
system have the legal status of private property, vital decisions about their use are made by
the individual businessmen who own them or by the corporate managers and government
officials who exercise the virtual rights of ownership. The complexity and obscurity of the
legal relations governing the use of our technologies, abetted by the reification that assigns
them to the realm of things—all of these help to create the aura of “phantom objectivity”
that envelops them.

Platforms are a technology subject to the same reification. Think of “Facebook:” one likely conjures
the user-interface accessed through a device, rendered in its signature blue and white and grey,
composed of profiles and newsfeeds and group pages. That is, one doesn’t first imagine Facebook as
a concatenation of social, economic and political relations at the production-level and the end-user
level. Critical platform scholarship, by examining the social, political, and economic context in which
platforms like Facebook operate, however, has done much to show how these sites and services are
shaped by and indeed composed of human relations. But this platform scholarship, with a few
exceptions (see Greene & Shilton, 2018; Kelkar, 2017; Shilton & Greene, 2019), has rarely broached
production-level platform work in practice. To rephrase Marx (2010/1997), this scholarship has
seldom looked to how ‘vital decisions about platforms use are made by the corporations who own
them and the employees that run them.’
The dearth of studies that closely engage with the operators of platforms is undoubtedly
influenced by challenges in gaining access to platform employees and executives, a population
famously skittish about talking to researchers or journalists lest these conversations violate signed
NDAs and risk their employment at that company (Roberts, 2019). This project thus benefits from
its focus on defunct platforms, where former employees were likely more willing to speak about

19
their work because it did not actively threaten their employment at the platform of interest.
1
By
conducting in-depth, semi-structured interviews with platform employees, and articulating the types
of worldviews that were expressed through employee and organizational practices during sunsetting,
this research provides a rare glimpse into how a platform’s ‘politics’ are socio-technically
constructed. This is illustrated especially in Chapter 2, where I show how corporate forms of
valuation shape the trajectory of a platform as it goes into decline, and in Chapter 3, where I
describe how employees decide how to handle user-data by drawing on their own personal values.
These are insights that can be mapped to platforms more generally, even ones that are not defunct.
This project thus attempts to further undo the ‘phantom objectivity’ of these technologies by
showing how these systems are socially and technically shaped–not autonomously conjured–over
time.
Digital Memory Infrastructure and Infrastructural Expectations
The dissertation does more than provide a critical and close reading of sunsetting as a
process. Bringing literature on sociotechnical infrastructures, social media platforms, and memory
and media together, it also argues that social media platforms are inseparable from the construction
and organization of the past today. Tracing the increasing cultural emphasis on media for purposes
of reflection, as well as the longstanding relationship between computational technologies and
human memory, I frame platforms as digital memory infrastructure in Chapter 1. Building on scholarship
that has argued for an understanding of platforms as infrastructural–that is, as sociotechnical
structures supporting modern society (Carse, 2016)–I provide evidence for understanding platforms
as computational systems that underly, support, and shape modern memory practice. In doing so,
this project contests dominant understandings of platforms as exclusively interested in presenting

1
For an in-depth discussion of mitigating participant risks and how other ethical aspects were considered in
this projects’ design, see Appendix A.

20
constantly refreshed, novel information to users. Instead, I argue for an understanding of platforms
and their business plans as intimately, if invisibly, tied to a human relationship to the past, especially
by virtue of their data storage.
Framing platforms as digital memory infrastructure, this study also contributes to
longstanding conversations about infrastructure in science and technology studies, and that
literature’s more recent intersection with the study of platforms (Plantin et al, 2018). In Chapter 5, I
analyze popular press accounts of sunsetting to make visible narrative tropes and thereby surface
latent cultural beliefs around platforms’ role as digital memory infrastructure. These narratives show
that there is an underlying expectation that content stored by platforms is stable and enduring. I
categorize these cultural perceptions of digital memory infrastructure as infrastructural expectations, or
normative beliefs about how an infrastructure should function. I highlight how infrastructural
expectations are inseparable from how an infrastructure functions in the social world: that is, what
one expects of infrastructural systems in general–whether they tend to be stable or faulty, leaky or
secure, useful or built simply for show–influences how one interacts with a given infrastructure. This
concept emphasizes how public expectations of digital memory infrastructure as stable and
permanent–an expectation that contrasts actual production-level platform practices, as I show in
Chapters 2, 3 and 4–creates conditions for greater cultural loss when these platforms shutter.
Sunsetting as Epistemic Practice
The concept of digital memory infrastructure encapsulates the idea that viable social media
platforms intentionally support–and indeed are expected to support–the organization of the past
through the maintenance of user-generated content over time. What this project examines is the
process wherein this support begins to fall apart, why it does, and what the effects are when it does.
The major consequence of this infrastructure coming apart is what I am calling commercial digital loss,
or the removal of user-generated content by the corporate channels hosting this content.

21
I provide evidence for this concept in Chapter 2, where I illustrate how platform companies
experiencing decline begin to economically (if not yet technically) pull apart pieces of the platform.
Rather than view the platform as a functioning whole, as they did when the platform was in a
growth stage, platform companies start to evaluate individual components of the platform–whether
its user audience, its patents, or its technical infrastructure–to determine which components have
extractable economic value. In this process of valuation, user-generated content is generally viewed
as possessing no value or even negative value to the platform organization. The inability to realize
commercial gains from user content thus frequently results in the removal of this content by the
platform company, or commercial digital loss. In Chapters 3 and 4, I examine the values that
platform employees bring to sunsetting, as well as the technical components of closure, showing
how these factors also help construct digital loss.
This project argues that commercial digital loss has epistemic effects, and therefore
sunsetting can be considered an epistemic practice. That is, how sunsetting proceeds shapes the
presence and absence of diverse mediated traces generated over time which come to represent that
past. Because mediation is increasingly central to modes of knowing and constructing the past
(Nora, 1989), as I show in Chapter 1, the presence or absence of mediated traces from platforms
stand to shape how this past is known and constructed over time. More specifically, sunsetting and
commercial digital loss are processes of constructing what historian Michel-Rolph Trouillot has
called “silences” in the historical record, which are embedded in sources, archives, narratives, and
historiography (1995). Deleting records of users––especially deleting records of those who would
not otherwise be represented in a formal historical record, who are historically excluded from
agentic representation, who are often representationally subordinated based on vectors of power
according to race, gender, and class––threatens to reinforce existing social hierarchies in the
construction of the mediated past. That is, in (re)producing silences, commercial digital loss

22
threatens to perpetuate “representational harms,” or harms that reinforce the subordination of
particular groups by virtue of identity (Barocas et al, 2017). In crafting these silences, sunsetting
represents another way that platforms come to bear on public discourse today. Through the
construction of destruction, they shape the past according to forms of value and valuation endemic
to the US techno-cultures examined in this project.
Key Terms: Platform, Content, and Sunsetting
Throughout this project, I use terms that have disparate and sometimes contested meanings. Here, I
clarify the use of three key terms around which the dissertation hinges–platform, user-generated
content, and sunsetting. This section serves two purposes: it first provides a bounded definition to
understand terms as they are used in this dissertation, just as it provides a critical foundation upon
which this research acts, expanding, shifting, or further clarifying these concepts.
Platform
There is a robust and growing body of research on platforms. A core debate in this research is
how to define platforms in the first place, and what analytical utility is found in any given definition.
This project adopts and builds on the basic definition of platforms that Tarleton Gillespie (2010;
2018) provides, understanding platforms as “online sites and services that [host], organize, and
circulate users’ shared content or social interactions for them” (2018, p. 14). Gillespie’s definition
could be characterized as a discursive stance toward the term ‘platform,’ which draws on evolving
popular definitions that refer to sites like Facebook, Twitter, YouTube, and others hosting user-
generated content as ‘platforms.’ Others have argued that platforms are defined by their
programmability, in that they are amenable to being programmed (Kelkar, 2017). These definitions
have their origins in corporate and technical circles (e.g. Andreessen, 2007), and encompass systems
like the Atari game console (e.g. Montfort & Bogost, 2009), that discursive stances would largely
understand as outside platforms’ definitional scope. While definitions of platform that foreground a

23
system’s programmability can be fruitfully mapped to the study of platforms hosting user-generated
content (e.g. Helmond, 2015; Helmond et al, 2019; Kelkar, 2017), programmability is not the
analytical focus of this project, while user-generated content is. Indeed, I build on Gillespie’s
discursive definition of platforms in Chapter 1, when I argue that platform research has largely
downplayed a core aspect of what it means to be a platform: the storage, hosting and organizing of
user-generated content over time.
Platform researchers have also recently argued for the productive overlaps between platform
studies and infrastructure studies (Kelkar, 2017; Nieborg & Helmond, 2019; Plantin et al., 2018;
Plantin & Punathambekar, 2019), a relationship this dissertation builds upon. For instance, Plantin et
al. (2018) have noted the ways that social media platforms like Facebook increasingly operate as
infrastructures, with characteristics of embeddedness, invisibility, and broad use. At the same time,
those systems typically thought about as infrastructures–from highways to the telecommunications
backbone–have started to resemble platforms, systems which are “privatized, fragmented, yet
interoperable” (p. 300). These definitional overlaps show how discursive definitions of platforms are
complementary to infrastructural characterizations, a generative overlap that this project builds on
through the sketching of digital memory infrastructure.
User-generated content
What is hosted by platforms and what is lost in commercial digital loss is “user-generated content.”
As Michael Daubs describes, user-generated content refers to (2020, p. 1826-1827):
…media content generated by people outside of professional media institutions, often for no
pay, which is made available to the public. This content can consist of text, photos, images,
graphics, audio, or video and can take many forms including consumer reviews, comments
on existing websites, posts on social networking platforms such as Facebook and Twitter,
memes, contributions to wikis such as Wikipedia, blog posts, podcasts, citizen journalism
contributions, or videos shared through sites such as YouTube or LiveJournal.

24
In this project, I at once employ Daub’s basic definition and address the ambiguity around what it
means for platform-based user-generated content to be ‘public.’ That is, the diverse media hosted by
platforms have gradations of availability to a public: on Instagram, one could share a photo to a
public account that is ostensibly viewable to everyone, though the platform could block those who
don’t have an Instagram account from viewing it; one could link to an article on Facebook but
adjust the default settings such that only some connections can see it; or one could simply send a
direct message between two people on Twitter. These examples have varying degrees of publicness,
but by virtue of their being generated by users, they are considered in this project to be user-
generated content.
The definition of user-generated content I employ here could, like platforms, be considered
‘discursive’ in nature, in that it is the term popularly used to describe varying media forms typically
hosted online, often by platforms. At the same time, ‘content’ has rightly been critiqued as being
ahistorical (Daub, 2020), lacking specificity (Andrejevic, 2009), or divorcing diverse media from the
morphology of their medium (Sterne, 2014). While content is used throughout this project because
of its succinct descriptive utility, I attempt to acknowledge these critiques in my use of the term.
For one, ‘user-generated content’ does not arise with the advent of digital media or online
platforms, even if the term became popular to describe user-made media in digital contexts (Daub,
2020). There is a long history of media production and media use outside of professional media
institutions, one that precedes the advent of digital technologies (Corry, 2020). While in the public
eye digitality has often been framed as utterly transforming the dynamics of media production,
sharing, and use, understanding user-generated content as superseding the exclusive domain of
digital computational technologies allows for a necessary historicization of media destruction and
loss. Just as media production has always also occurred outside of media institutions before the
ubiquitous use of computational digital technologies–whether the production of zines or scrapbooks

25
or pamphlets or letters–so too has large-scale media destruction occurred prior to their use too,
whether the intentional burning of books to control a public’s access to information, the destruction
of archives through environmental disaster, the tossing of a shoebox that contained stacks of
photos, or the shredding of documents to hide corporate malfeasance. What this project reveals
then is not the first ever time media content has been destroyed, but rather the contours of cultural
loss as they are (re)configured on social media platforms.
Moreover, the use of ‘content’ as a catchall term to refer to both media produced by users as
well as the data generated by users in their online interactions, which is then commoditized (e.g.
insights about browsing habits that then help ad targeting), threatens to dull analytical specificity
(Andrejevic, 2009; Hesmondhalgh, 2010). In this project, I employ user-generated content to refer to
the diverse media actively made by users, while user-generated data refers to all traces left by users.
However, because user-generated content has commodity-potential (explored in Chapter 2), it is not
analytically accurate to divorce them completely, and thus when I refer to user-generated data it
encompasses content under this umbrella. When greater specificity is needed between these terms in
this project, I offer additional clarification.
Finally, discussing content as a standalone entity risks divorcing the ‘message’ from the
‘medium,’ to use Marshall McLuhan’s terms (1964), or the ‘content’ from the ‘form,’ as Jonathan
Sterne has phrased it (2014). This separation may frame content as substantive stuff that travels over
or fills up a neutral technical infrastructure (ibid), an analytical stance which may fail to critically
engage with both production-level power and the political socio-materiality of these systems
(Winner, 1980). By focusing on content as it is handled at the production-level of platforms, I hope
to have avoided this oversight.

26
Sunsetting
In contrast with the terms explored thus far, sunsetting is one that has not been up for
scholarly debate. Sunsetting is used colloquially within technology industry contexts to refer to the
closure of a product, site or service. An etymological tracing of ‘sunset’s’ verb form suggests that
this use originated in US governmental practice, referring to legislation stipulating that a program or
measure must be terminated at the end of a fixed period unless it is explicitly renewed (‘sunset,’ v.,
1976). Ensconced in ‘sunset clauses,’ ‘laws,’ or ‘provisions,’ it is the acknowledgment–nay, the rule–
that something will end.
In contrast with the governmental use of sunset, sunsetting in platform contexts is less
clearcut and typically less planned. In an article drawing on interviews from this project (Corry,
2021), I show how a platform’s death is conceptualized heterogeneously by platform employees.
That is, a platform’s death is not caused by a universally agreed upon-set of events, nor is there a
clear moment at which ‘death’ happens: it is not simply the pulling of a plug. Instead, a platform’s
death is negotiated and defined by platform employees within the context of the high-tech industry,
as employees use stories about a platform’s death to attest to their own industry knowledge within a
neoliberal marketplace. Like platform death, sunsetting does not always have a clear start and end
point. It is a process of closure, a process influenced by actions deliberately taken toward shutting a
site down, but also one influenced by social and technical decisions made before the period of active
shutdown, as I show in chapters that draw on interviews.
At the same time, it can be unclear if a platform could even be considered sunset (if we
consider sunset here as an adjective, instead of a verb). For instance, is MySpace a sunset platform?
One can still access the MySpace URL, after all, even with its 2019 content loss. Yet I consider this
platform as one that is sunset. That is, if platforms are defined by their ability to host, organize, and
circulate user-generated content, as I characterized earlier, then a platform has been sunset if it no

27
longer supports those actions. Sunsetting, as it is explored here, can therefore be considered the act
of constructing–or socio-technically shaping–a state of destruction for digital memory infrastructures.
Dissertation Structure
Sunsetting: Platform Closure and the Construction of Digital Cultural Loss is composed of five
chapters and a conclusion, as well as two appendices: one describing this project’s methods
(Appendix A), the other providing a sample of interview questions (Appendix B).
Chapter 1, “Memory, Media, Man, Machine: A Genealogy of Digital Memory
Infrastructure,” shows how social media platforms have built on a longstanding cultural alignment
between computational media storage and user memory, so much so that they act as digital memory
infrastructure. I define digital memory infrastructure as computational systems that underly, support,
and shape modern memory practice, and in doing so, argue that platform systems continue to act as
structuring forces for the past even as they decline and sunset.
Chapter 2: “Vicissitudes of Value: The Commercial Logics Behind Digital Loss,” focuses on
interviews with Friendster and MySpace employees to trace processes of economic and cultural
valuation as a platform goes into decline. It shows how in moments of decline, notions of what has
value and where value resides on a platform start to shift. In particular, I illustrate how user-made
media hosted by the platform can move from an asset to a burden from the perspective of the
platform company, revealing the commercial underpinnings of commercial digital loss.
In Chapter 3: “Social Values and the Shaping of Platform Afterlives,” I draw on interviews
with employees from MySpace and Vine to show how personal values, or what someone deems
important in life, come to shape sunsetting. I focus on employees’ attitudes toward user-generated
content and memory to describe how employees come to make decisions about what to do with
user-generated content. Chapter 4: “The Construction of Destruction: Material Considerations in
Sunsetting” continues the story of MySpace and Vine to describe how the technical components of

28
a platform also play a role in shaping the magnitude of commercial digital loss. I conclude these
empirical chapters by reflecting on the ways that destruction is actively constructed through
intertwined social and technical vectors.
Chapter 5: “Sunsetting, a Story: Media Accounts of Platform Closure,” zooms out from
production-level practices to understand narrative tropes in popular English-language press about
platform sunsetting. Analyzing press from the sunsetting of GeoCities, Friendster, MySpace, and
Vine, I show how dominant perspectives view platforms as relatively stable repositories for user-
generated content. Phrasing these perspectives as ‘infrastructural expectations’ that are brought to
digital memory infrastructure, I contrast these expectations with the production-level practices on
display throughout the project.
The Conclusion reflects on how, during sunsetting, platform organizations and employees
engage in speculation about the value of platforms, their users, and their content. I ask about other
ways that the long-term value of these mediated traces can be speculated upon, whether through
their potential for personal memory, historical knowledge, or artistic reflection. I ultimately reframe
sunsetting outside of the social media platform context, and argue that analysts should foreground
acts of deletion and removal as salient nodes for understanding power, sociality, and materiality in
information systems today.

29
Chapter 1
Memory, Media, Man, Machine:
A Genealogy of Digital Memory Infrastructure

In September of 2011, Facebook announced a fundamental reworking of the site’s design at
their annual developer conference, F8. Company founder Mark Zuckerberg—27, t-shirt, sneakers,
jeans, headset—paced the unadorned stage as he described “the heart of your Facebook
experience—completely rethought from the ground up” (CNN, 2011). The company, which had
cemented its status as the leading profile-based social network in the United States, was largely
considered to be defining what social media platforms would, and could, be (Miller, 2011). Moving
from user profiles that were essentially a calling card and message board––functions not so different
from those available since the pre-web days of bulletin board systems––Facebook profiles would
now be organized through a system they dubbed ‘Timeline.’ Timeline, as he described, was “all your
stories, all your life and a new way to express who you are.” The redesign created a master narrative
about you and your past, browsable by you and your network. The individual profile was turned into
an easily navigable chronology, divided by year, on which algorithmically selected important events
were shown, either through content that the user had already posted on Facebook or chose to
include retroactively. As Zuckerberg would repeat throughout his introduction to the feature: “It’s
how you can tell the whole story of your life—on a single page.”
Facebook’s earliest users had joined the platform as college students in 2004 and had been
on the platform for seven years when Timeline was announced. These years, for some, constituted
the transition out of college and into major life events: new jobs, locations, relationships, marriages,
pets, children—events that were in some way chronicled through content that had been uploaded to
the site. Even users who had joined in 2006, when Facebook was opened to those outside of the

30
colleges and corporations they first welcomed, had five years of content hosted and organized there.
As a New York Times article describing the change noted (Pogue, 2011):
…for regular Facebookers, the Timeline serves a real purpose. For example, if you got
engaged a few months ago, only the Facebook regulars among your fans might know it. Oh,
they could keep clicking More, More, More, to summon older and older posts you’d made
— but how would they even know to do that? Now there’s a way for them to see the arc of
your life in a visual, entertaining way — a genuinely useful online tool that nobody’s carried
out in quite this way before.

Timeline responded to the abundance of content that users had uploaded there by winnowing what
could be easily seen about a user’s past engagement. But it also acted as a strategic play for its users’
enduring attention and therefore the platform’s long-term persistence. After all, social media
platforms did not begin with the express desire to be used for memory. But the storage of
meaningful media, organized in such a way that evokes the story of one’s life and one’s social
interactions over time, positioned the platform as an indispensable maintainer of valuable personal
information—that is, the maintainer of “the whole story of your life.” As the business scholar
Catherine Tucker has argued (2019, p. 688), technology users are discouraged from switching to
other platforms when the benefit “that the user derives from using the platform primarily lies in the
fact that the platform stores their historic data for them,” making for high switching costs and thus
becoming a source of the platform’s market power. Facebook’s Timeline was a play to become
indispensable to users through the conscientious organization of the personal past.
More specifically, Timeline’s introduction can be understood as a strategic move toward
becoming infrastructural: infrastructure for making sense of the past; infrastructure for memory.
2
As

2
Ten years after Timeline was announced, the company announced another fundamental shift that can be
considered an ultimate move toward becoming the infrastructure of users’ lives. Rebranding their suite of
services—including verticals like Facebook, Whatsapp, Instagram–under the name ‘Meta,’ the company also
added an integrated virtual world they dubbed the ‘Metaverse.’ An hours-long tour of the Metaverse framed it
as a virtual realm in which all of life could and would unfold--work, play, sociality, and more. The
introduction of Timeline as a way to tell the ‘whole story of your life on a single page’ can be understood as

31
platform scholars have argued, sites and services like Facebook increasingly act as infrastructures in
myriad ways. Showing common infrastructural characteristics like ubiquity and embeddedness while
supporting the basic functions of modern society, these interlocking systems underly education,
sociality, the economy, transportation and beyond (Helmond et al, 2018; Kelkar, 2017; Plantin et al,
2018; Plantin & Punathambekar, 2019; Van Dijck et al, 2018). In understanding these entities as
infrastructure, and showing how platform companies often pursue infrastructural ambitions
(Helmond et al, 2018), this research has shed light on contemporary transformations in society,
especially how platforms intervene in the organization of public, democratic values (Van Dijck et al,
2018) and the ways that infrastructural systems built for public interest have increasingly been
replaced with platform systems organized around private gain (Noble, 2018; Plantin et al, 2018).
Building on this literature, I argue in this chapter that platforms also constitute what I call
“digital memory infrastructure,” or those computational systems that underly, support, and shape
modern memory practice. Thinking about memory as having an infrastructure is what
anthropologist Brian Larkin (2013) has deemed a “categorizing moment,” in that identifying
something as infrastructural provides a “cultural analytic that highlights the epistemological and
political commitments involved in selecting what one sees as infrastructural (and thus causal) and
what one leaves out” (p. 330). This categorization indeed requires some abstraction–it is less
concrete (often literally) than other artifacts commonly called infrastructure, like roads and tunnels,
sewage pipes and telephone wires. However, understanding contemporary memory practices as
being supported by an infrastructure draws attention to the ways that computational systems’ socio-
materially shape the organization of the past, and serves as a foundation for understanding
platforms’ important and value-laden role in this process.

establishing a foundation for the more ambitious project proposed in 2021, a transhumanist vision in which a
whole life could be lived out on a single platform.

32
Facebook’s Timeline is just one part of digital memory infrastructure’s story, a narrative that
consists of evolving social theories of memory, changing forms of mediating and of organizing the
past, and persistent cultural imaginaries around the relationship between computers and memory.
This chapter charts this evolution to illustrate the shifting relationship between these concepts over
roughly the last century of western thought. By charting this relationship, I show the process
whereby computational media have become intractable, embedded, and ubiquitous elements in
constructing the past.
In what follows, I discuss canonical literature from the multidisciplinary field of memory
studies, illustrating the intensification of media use in discussions of what memory is and how it
functions over the last century. I then turn to the ways that the western cultural imaginary has long
understood, and built toward, the use of machines—especially computational machines—for
organizing and making sense of the past. I extend this discussion to show how platforms are part of
this computational trajectory, incorporating memory functions as inseparable elements of their
architecture, and in turn argue that platforms are defined in part by their ability to mediate the past. I
thus extend the history of memory and machines in this section to platforms while also showing the
ways that platform companies have made strategic plays to be the invisible supporter of individual
and collective memory, key elements in digital memory infrastructure.
Theories of Memory and Media in the 20
th
Century
Has mediation always been inseparable—in theory and in practice—from memory?
Literature from memory studies, a multidisciplinary field under which the intertwined social,
political, cultural, technological and psychological dynamics of remembering and forgetting are
studied, suggests that they have not. The history of memory, the past of organizing the past
collectively and individually, is marked by a range of tactics, ideologies, and uses. Many do not
involve mediation in the traditional sense of having a material conduit through which information is

33
transmitted. For instance, art historian Frances Yates’ (1966) influential early writing on memory
practices in the European Middle Ages and Renaissance shows the diverse ways in which the titular
“art of memory” was practiced. The most famous and long-running of these modes of memory—
practiced from ancient Rome through Middle Ages–was the construction of a “memory palace,” and
the trained association of certain thoughts with certain rooms in that imagined location. As she
describes (Yates, 1966, p. 4):
In the ancient world, devoid of printing, without paper for note-taking or on which to type
lectures, the trained memory was of vital importance. And the ancient memories were
trained by an art which reflected the art and architecture of the ancient world, which could
depend on faculties of intense visual memorisation which we have lost.

Along with these highly formalized memory practices in Europe, the past was otherwise defined for
many through oral culture. Moreover, the cultural relationship to the past was characterized not by
teleological distance but by a sense that the Christian, religious past was essentially alive in the
present (Nora, 1989; Yates, 1966). Even with the debut of media technologies like the printing press
and lithography, the ancient art of memory continued to be practiced, and was even further
entrenched as a mode of remembering as printed media emerged (Bolzoni, 2021). That is, memory
was not always entangled with mediation, and accounts like Yates’ (1966) show that memory
practice did not always want for nor require mediation. Nonetheless, as I discuss in this section, as
the field of memory studies developed over the course of the 20
th
century and looked to memory
practices of the 20
th
century, media use became an increasingly important part of conceptualizing
what memory was and how it was practiced.
The beginning of memory studies as a field is typically cited as the publication of the French
Durkheimian sociologist Maurice Halbwachs’ The Collective Memory (1925/1992). I begin this
discussion of media and memory here to offer a starting point for thinking about the gradual
emphasis on media technology use in recollection, both within memory studies but also within

34
general social practice. To put it simply, the primary contribution of this origin point was in arguing
for memory as a social function, not a purely psychological faculty. Halbwachs discusses how
collective memory frameworks work within a social group–that of the family, of religion, and of
professional activity. Memory, then, is formed through a dialectical process that goes between the
individual unit and the society-level group: there are social processes in each specific group (for
instance, my family) derived from each general group (for instance, family in general) that determine
habits of remembering and forgetting. Media and technology, that is, are far from a focus, and
largely exist in the background. For instance, Halbwachs considers the ways that elders play the
social role of maintainers and communicators of the past within families (and indeed across
societies). One of the ways they engage in this role is by revisiting and creating media: elders “go
through old papers, old letters; above all, they tell what they remember, if they do not try to write it
down” (p. 48). Media technologies are not some stable bearer of meaning that transmutes the past
wholesale to the present; rather, these traces are always reanimated around the present needs and
dynamics of the social group.
As the study of memory as a social phenomenon evolved in the 20th century, the focus on
media technologies for memory practice intensified. Pierre Nora’s Lieux de Mémoire (1989), a central
point in memory studies eclipsed only by citations of Halbwach’s work, provides evidence of this
gradual shift. Building on theories of social memory, Nora engages in a meta-historical mapping of
memory’s function in French society. He bases this study in two concepts: lieux de mémoire, which
describe sites of memory invested with symbolic memorial weight, and milieu de mémoire, which
describe “real environments of memory” that are social, continually activated and most at work, in
Nora’s treatise, in “primitive” or “archaic” societies (p. 8). These milieu de mémoire describe societies
where the past is contiguous with the present; historical societies, in contrast, see the past as

35
disjointed from the present, the present always moving further away in its relationship to what came
before.
Nora makes the postmodern argument that these lieux de mémoire, whether statues,
anniversaries, national holidays or beyond, are vestiges of real memory environments in an ever
more historical, modern society. That is, when real memory is overtaken by a historical
consciousness driven by an imperative of change and progress, lieux de mémoire are created as
remnants of a “memorial consciousness” that would (p. 12), without commemoration, be swept
away by the forward march of history, by modern societies’ taste for newness. To put it differently,
if a society fully adopted a historical consciousness, no traces of the past would be preserved at all,
no lieux de mémoire would exist.
Traces, archival records, memory institutions, and historical accounts–media, in other words–
are primary evidence of this shift, Nora argues. As he writes, in milieu de mémoire, “each gesture, down
to the most everyday, would be experienced as the ritual repetition of a timeless practice in a
primordial identification of act and meaning” (p. 8). Yet “with the appearance of the trace, of
mediation, of distance, we are not in the realm of true memory but of history” (p. 8). While
mediation might seem to be co-constitutive of this shift–each new record an erosion of timeless
practice–Nora understands increasing mediation to be a result of this shifting consciousness, not a
cause of it. Like Halbwachs, Nora argues that changes in how we relate to the past, and how we
mediate the past, come from overall structural changes in society: in his case, French
industrialization and the erosion of peasant culture. At the same time, the primacy of media in
Nora’s account is evidence of a growing focus on the entanglements of media technologies,
materiality, and social recollection, just as his historical account attests to the growing use of material
artifacts for providing continuity with the past.

36
Toward the latter half of the 20
th
century, scholarship emerged that positioned media as a
primary means of sustaining the past through time, largely through the concept of cultural memory.
These texts focused on important mass cultural events over the course of the 1960s through 1980s;
Marita Sturken’s (1997) formative work on cultural memory focuses the Kennedy assassination, the
AIDS crisis, and the Vietnam War. As Jan Assmann and John Czaplicka write (1995), cultural
memory comprises that “body of reusable texts, images, and rituals specific to each society in each
epoch, whose ‘cultivation’ serves to stabilize and convey that society’s self-image” (p.132). It is
“exteriorized, objectified, and stored away in symbolic forms that, unlike the sounds of words or the
sight of gestures, are stable and situation-transcendent. They may be transferred from one situation
to another and transmitted from one generation to another” (Assmann, 2008, p. 111). Moreover,
cultural memory typically operates outside of formal historical discourse, yoked instead to popular
cultural discourses (Sturken, 1997). Cultural memory research thus builds on social understandings
of memory as being activated and formed in social groups, but emphasizes the ways that
objectivization, externalization and mediation allow pasts to persist in that group in part because of
material stability. Texts engaging with cultural memory, in turn, have examined a variety of
objectivized, mediated symbolic cultural forms, whether literature and photographs (Hirsch, 2008),
souvenirs and keepsakes (Boym, 2001; Hartman, 2007), Hollywood films, amateur videos,
performance pieces, and folk art (Sturken, 1997), or newspapers and scrapbooks (Pollard, 2012).
As these scholars show, cultural memory does not begin at the moment when an object from
the past is interpreted in the present. Rather, groups engage in cultural memory work, creating
objects they believe should or will exist past the moment of recording. Zelizer (2002), for instance,
shows how chronicling traumatic events through personal photography–whether Allied soldiers’
images taken after the liberation of Nazi concentration camps in 1945, or photographs of the World
Trade Center site in the weeks following 9/11–is an active process of bearing witness which

37
anticipates these images’ future cultural use. These practices also show the ways that cultural
memory is shaped by culturally specific understandings of media as important bearers and
circulators of the past. Photography, for instance, was chosen as the appropriate means of
recognizing these events as they unfolded. Cultural memory is entangled with media, then, in both
practice and in product.
These objectivized, externalized, or mediated cultural forms shape the activation of the past in
the present. They structure ideas about the past not just through symbolic testimony–whether a
souvenir from Disneyland which figuratively declares “I went there” or a photograph of Ground
Zero that shows “this did happen”–but through their ability to circulate among social groups,
relating individual to collective, crafting unified cultural ideas about the past in this process
(Landsberg, 2004). A particularly salient means of understanding how media’s evidentiary quality in
combination with its ability to be circulated shape ideas about the past is through cultural memory
theorists’ articulation of “screen memories” (Sturken, 1997). Originally coined by Freud to describe
individuals’ “screening out” of painful memories, screen memories have also come to be
characterized as memories which are shaped by mediated representation–by screen images (Sturken,
1997). For instance, though the recording of JFK’s assassination was initially released and publicly
shown in 1975, some recount seeing JFK’s assassination live on television in 1963 (Sturken, 1997, p.
28). Screen memories show the slippage, and in turn the inextricability, of media and memory. Yet
cultural memory scholars don’t take these screen images as rote, all-powerful rewritings of memory
through mediation; rather, the continual use of these images–the emergence of new stories, new
contexts, and new cultural discourses around them–always may reshape what these images say about
the past.
Cultural memory theorists therefore understand cultural objects, media among them, to be
means by which individuals and groups articulate, assert and perpetuate specific meanings of the

38
past. Cultural memory work shows the ways that circulation, access, and persistence of objects shape
our interpretations of what came before. Moreover, while cultural memory definitions might
understand material persistence as one component of cultural memory’s stability (Assmann, 2008),
work in this domain generally understands media and memory as mutually shaping. The notion of
cultural stability created through exteriorized, objectivized, and mediated pasts, therefore, is as much
cultural as it is material. Or, as Sturken puts it (1997), cultural memory’s power is “derived not from
its revelation of any original experience but from its role in providing continuity to a culture, the
stakes in creating values in that culture, and the fundamental materiality by which that culture is
defined” (p. 259). In contrast to those core theories of social and collective memory, cultural
memory recognizes the ways that the mediation of memory is not just a side-effect of larger social
processes, but rather an active element in constructing the past.
In the last decade of the 20
th
century–and continuing into the new millennium–‘the archive’
emerged as a popular theoretical construct for understanding the truth claims of documentation
(Stoler, 2002). The ways that records are physically embodied, collected, organized, accessed, and
used are all epistemic considerations, shaping what we may know. The archive, in turn, became an
epistemic project and a powerful metaphor for remembering and forgetting (Stoler, 2002), linked to
both formal historical discourse as well as more vernacular forms of memory. This archival turn also
marked the opening up of archives as a site of theoretical and scholarly investigation, prompting
renewed consideration of institutional archives as non-neutral sites of power with considerable
influence in shaping historical understanding (Schwartz & Cook, 2002). Crucially, this epistemic
symbol was also linked to the emergence of digital media.
This archival focus was prompted in part by Jacque Derrida’s 1995 “Archive Fever: A
Freudian Impression,” adapted from a lecture given at psychoanalysis conference Memory: The
Question of Archives. “Archive Fever’s” articulation of ‘the archive’ signaled a shift in scholarly

39
understanding of remembering and forgetting, one in which what we can know about the past is
strongly determined by media’s technical and material characteristics. Derrida articulates this
reframing in part through a “retrospective science fiction” about the field of psychoanalysis, asking
if the means of recording had been different–if Freud and his colleagues had access to emails and
computer, televisions and credit cards–whether the field itself would look anything like it did the day
of his lecture. As he asks, had the “whole of this field been determined by a state of the technology
of communication and of archivization?” (p.17) Archivization–those technical means of capture and
storage–determines more than what makes it into the archive, he argues. Rather, these technical
means determine what an event is in the moment of its happening, and what it becomes, how it may
shape the future. As he puts it: “the technical structure of the archiving archive also determines the
structure of the archivable content even in its very coming into existence and in its relationship to
the future. The archivization produces as much as it records the event” (p. 17).
While Derrida does not treat memory’s construction as explicitly as those collective, social, or
cultural memory analysts discussed so far, his assessment of what we come to know through
technical aspects of mediation speaks to broader considerations of how we know the past, on both
individual, group, and larger collective scales. Derrida lays the foundation here for sustained analyses
of mediation as a primary driver of memory, instead of media understood as something that co-
produces memory or is subordinate to it. As such, this text and the vectors of both memory and
archival studies that it spawned attests to the intensification of media and memory entanglements
just as the networked digital world was making its popular debut.
As these cases and related scholarship from memory studies show, the intensification of media
use in the production, maintenance and conception of the past over the course of the 20
th
century
has arisen through the debut of novel media forms and the increasing accessibility of these media
forms (Landsberg, 2004; Van Dijck, 2007), along with broader socio-cultural shifts that prompted

40
changes in how the past was organized and related to (Nora, 1989; Yates, 1966). Together, these
texts show how media came under increasing scrutiny as vectors for memory, just as they also speak
to the increasing cultural emphasis on media for organizing, sustaining, and shaping the production
of the past. They highlight the intensification of the entanglements between memory and mediation,
or differently, the growing inability to disentangle memory practice from media technology use. In
doing so, they provide a foundation for understanding contemporary memory as having an
underlying, mediated infrastructure.
Machines and Memory at Mid-Century
Just as memory studies scholarship was observing how mediation shaped a relationship to the
past over the course of the 20
th
century, technologists and the broader public in the same period
imagined the ways that emerging digital computers interfaced with memory. At mid-century,
computers (both analog and digital) developed during World War II were publicly celebrated as
technological marvels that had aided in the Allied forces’ military success. These devices could
crunch numbers at unprecedented speeds, control ballistics, and crack otherwise impenetrable
enemy ciphers. Computers also rapidly entered mainstream consciousness with more flexible uses,
heralded for being able to forecast the weather (many thought the next step would be to control it
[Alarcon et al, forthcoming]) or predict elections, evinced when the Univac successfully predicted
the presidential election winner in 1952, having calculated it using just early returns from the
national vote (Draper, 1953).
In this rich cultural imaginary surrounding what computers could and would do, these devices
were early on associated with what they would and could do for memory. As I’ll discuss, these
imagined memory functions tended to differ from the personal and collective iterations of memory
discussed by memory studies scholars, instead emphasizing memory as a scientific faculty useful for
organizing and relating to accumulated scientific knowledge. Nonetheless, associations between

41
computing and memory would likewise lay the groundwork for the contemporary relationship
between digital communication technologies and social memory functions of today. Moreover, the
type of memory practice that was imagined–what the useful past constituted, and why computers
were crucial elements in mediating it—reflected ideological and epistemological positions very much
shaped by vectors of power that persist today. Here, I discuss early associations between memory
and computing in the mid 20
th
century to bring out these understandings, focusing in particular on
the memory device proposed by American computer scientist Vannevar Bush––the ‘memex’––as
well as the primacy of memory functions in early popular analogies of computers as ‘brains.’
The Memex
The most prominent example of thinking at this intersection was the memex, an imagined
device first introduced by American computer scientist and former director of the American Office
of Scientific Research and Development Vannevar Bush in his 1945 Atlantic essay, “As We May
Think.” The memex—and the essay itself, which introduces a number of other futuristic gadgets—
has been heralded as a prescient vision of what computing and especially the internet would be
(Schrage, 1995). It has also been well-cited within media and memory studies scholarship, its
prescient quality often critiqued (e.g. Chun, 2011; Leslie, 2020). This chapter engages these well-
established grooves to illustrate and argue that the memex was an important conceptual
development in the longer history of digital memory infrastructure. The memex is not important
because it foresaw a specific computational future, but rather that the attention given to it provides
evidence of the prominence of memory functions in early imaginaries around personal computer use
and, eventually, around a broader public’s use of the internet.
The memex–whose abbreviation is up for debate, comprising either a memory index or memory
extended (Yeo, 2021)––used recent developments in document compression and miniaturization,
like microfilm, to extend the personal memory of an individual. The memex was a device for

42
information organization and access, just as it was a machine for freezing and recording associations
of thought between bodies of knowledge and pieces of information. As Bush (1945) writes in “As
We May Think:”
Consider a future device for individual use, which is a sort of mechanized private file and
library. It needs a name, and, to coin one at random, “memex” will do. A memex is a device in
which an individual stores all his books, records, and communications, and which is
mechanized so that it may be consulted with exceeding speed and flexibility. It is an enlarged
intimate supplement to his memory.

The memex’s true innovation, in Bush’s thought, was not necessarily the miniaturization,
compression, and storage of information. Rather, it was the freezing of “associative trails” of
thought. Its main contribution was in including:
…associative indexing, the basic idea of which is a provision whereby any item may be caused
at will to select immediately and automatically another. This is the essential feature of the
memex. The process of tying two items together is the important thing.

Some have argued that the memex was crucial for inspiring the development of numerous
ubiquitous technologies, from hyperlinks to the internet itself. The memex has especially taken on a
hallowed place in internet lore since the late 1990s, because it seems to track closely with the way
that information is stored, organized, accessed, and retrieved using the web (e.g. Allington-Smith,
2015). As Wendy Hui Kyong Chun has noted (2011), this idea has been solidified in large part
because ‘luminaries’ of the internet age, like Douglas Englebart and Ted Nelson, have cited the essay
as an important touchstone for their own ideas and inventions in computing and the internet. That
the memex looms large in stories about the development of the internet at once points to
teleological historical thinking about technology, just as it signals a fundamental association between
perceptions of the internet’s potential and memory practice.
Yet memory here, as I noted earlier, does not seem to constitute the affective, socio-cultural,
or personal memory that memory theorists of the 20
th
century tended to address. While

43
conceptualized around memory recall and mediation, it is not the kind of memory recall associated
with mnemonic media like photo albums, yearbooks, and mementoes. Instead, memory for Bush
meant access to knowledge. Better access to knowledge relied on the storage, organization and rapid
recall of a comprehensive store of information. The memex was envisioned as the technological
medium through which this type of memory could be realized. Bush, much like Marshall McLuhan
argued two decades later (1964), understood advances in technology as having allowed humans to
supersede their natural senses; sight was extended through telescopes and microscopes, hearing was
extended through radio and telegraphy, and so on. The memex, however, represented an extension
not just of an individual human faculty but of human thinking through its relationship to memory.
Bush conceptualizes the memex as extending the faculties of thought in two ways: allowing humans
to take in more information than they otherwise could in a situation conceptualized as suffering
from ‘information abundance,’ and in extending the faculties of recall for what are otherwise
transitory thoughts.
This abundance of information was in large part a product of the recently constructed US
military industrial complex, a melding of military, industry, and academic researchers who typically
functioned with little oversight. Bush had been a primary architect of this system, convincing then-
president Franklin Delano Roosevelt for funding and political support in 1940, creating the National
Defense Research Committee (Wardrip Fruin & Montfort, 2003; Zachary, 1997). As historian
Daniel Rosenberg has described (2021), Bush imagined the memex as a response to the perceived
proliferation of scientific knowledge of the post-World War II era, a means of remembering and
navigating this knowledge and the memex user’s own relation to that knowledge. As Bush notes
(1945), using the associative trails afforded by the memex, the user “builds a trail of his interest
through the maze of materials available to him. And his trails do not fade.”

44
Whose memory—whose access to knowledge—does the memex mediate? The imagined users
are, in a phrase, “men of science.” In the vision of “As We May Think,” the memex is used for the
personal organization of knowledge and provides a means for the individual to orient (him)self
toward the stored accumulation of all human knowledge. While not a communal device, an
individual might use it, as in Bush’s example, while talking to a friend about political or scientific
topics–like the historical avoidance of technical innovation. The memex also embodies a particular
global-individual relation, positioning the user as able to master and classify the world’s knowledge
across time. Crucially, the memex is not a memory device of the home, to be placed in the living
room next to the television; it is both placed and gendered far from the symbolic feminized realms
of the home. When women appear with machines in his essay, they are showgirls, for instance
“languidly stroking the keys” in public displays of the recently invented vocoder. We may also be
safe to assume that these users are largely white and American, as essay illustrations, imagined
scenarios, and the demographic of most “men of science” at that time would seem to signal.
By envisioning a future for memory significantly tied to the memex, “As We May Think”
positions memory as significantly machine-mediated and highly epistemic in nature. Memory
without the memex relies on inevitably faulty, forgetful human faculties; the past, useful for its
accumulated knowledge, is therefore accessed through devices capable of ‘extending’ and perfecting
faculties of recall. As Bush notes, the memex would make it “possible to beat the mind decisively in
regard to the permanence and clarity of the items resurrected from storage.” It is a type of memory
valued because it provides fodder for scientific innovation. What this essay also illustrates are the
ways that imaginaries of machine-aided memory from the early years of computation, cited as highly
influential within American technological cultures, are value-laden. While later visions for machine-
mediated memory would adopt a falsely neutral ‘memory for all’ ethos (a position I discuss in detail
in Chapter 5), “As We May Think” shows very clearly how machines for memory encompass

45
particular visions of memory, particular visions of who the past serves, why it is valued, and how it
should be recalled.
The Electronic Brain
The idea that a machine could “extend” the brain, perfecting its faculties, is not limited to the
vision of the memex. Indeed, in the 20
th
century, early computers were conceptualized by the
American and British press as “electronic,” “mechanical,” “magic,” or “giant brains.” This metaphor
became prominent toward the end of World War II, when news outlets were able to discuss these
machines openly––for national security reasons, press had been limited during the war (Martin,
1993). These accounts heralded ‘electronic brains’ as allies in bringing the war to an end (ibid). In the
earliest years of these fantastic and anthropomorphized visions of computers, these brains were just
as likely to be digital computers like the ENIAC as they were to be flight regulators (“Electronic
Brain Records,” 1944) and projectile simulators (Gotthart, 1944) powered by analog computers.
Popular press accounts relying on the metaphor of the brain were often hyperbolic, fueling
speculation that these brains could outdo human thinking (Martin, 1993).
Scholarly work parsing early visions of computing has examined this metaphor. Computer
scientist C. Dianne Martin (1993) has noted how the dominant vision of the computer in the early
1960s was of an “awesome thinking machine,” in large part due to press from the late 1940s through
the 1950s. Design historian Paul Atkinson (2014) has similarly argued that the presentation of
computers to the general public in the 1950s cemented the idea of computers as brains capable of
solving previously intractable problems, a vision that both journalists and computer manufacturers
eagerly responded to.
What becomes clear in accounts like Atkinson and Martin’s, that look to press use of this
metaphor, are the faculties that made computers appear to be brain-like, though these attributes are
not explicitly discussed in their writings. I review the sources presented in their work to show that

46
memory was an important early element that helped construct the imaginary of the computer as an
intelligent machine. (It is a metaphor that would be taken up by cyberneticians in the ensuing
decades, though their analogies were more so inspired by advances in neurobiology [Bowker, 2006;
Medina, 2011]). In doing so, memory would become indelibly tied to the popular image of the
computer and to imaginaries of how computing would interact with the social world.
One of the earliest accounts in American newspapers was a 1941 piece in the Des Moines
Tribune, chronicling John Atanasoff and Clifford Berry’s work on the ABC Computer at the
University of Iowa. A particularly early example of computers being discussed in the press, Martin
(1993) highlights its use of the electronic brain metaphor. Crucially, however, what the article homes
in on as the key feature of this electronic brain was its capacity for memory. “Machine Remembers,”
the article is titled, with a picture of Berry and the computer’s memory unit. As the article noted,
“An electrical computing machine said here to operate more like the human brain than any other
such machine known to exist is being built” [7, box 145].
Articles written after World War II about computers like the ENIAC focused on these
machines’ ability to solve problems, solve them fast, and do so by virtue of memory, as articles like a
1946 London Times piece “An Electronic Brain: Solving Abstruse Problems; Valves with a Memory,”
would evince. A 1947 article about the Mark I and II at Harvard likewise foregrounded these
qualities:
Like the brain, the machine accepts information, generally in the form of figures represented
by small holes punched in paper tape. It salts them away in a kind of ‘memory.’ (Dr. Aiken
prefers the relatively modest term: storage of numbers.) Then it combines them into
conclusions, as human brains try and often fail to do. Unlike most human brains, it stops
when it makes a mistake.

A 1947 article in the Philadelphia Evening Bulletin put it similarly:
At the will of the operator, the tube recalls and applies to new equations any required
combinations of the “remembered” problems—at any stage of their solution. By turning a

47
dial, the operator can drop the tube’s “memories” back into a kind of electronic
subconscious- -or he can make the tube “forget” them entirely ...If an operator, feeding in
problems on perforated tape, has made a mistake, the machine will tactfully ring a bell!

Atkinson’s discussion of anthropomorphizing accounts (2014) equally features memory as an
important component in understanding computers, and understanding them as brains. Indeed, a
crucial development in computing was the addition of magnetic tape memory drives, which were
cheaper than previous data storage options, offered unlimited offline capacity, and were
standardized, making them interchangeable (Computer History Museum, n.d.). Rather than being
hidden away deep inside the device, these tapes were visible from the outside in glass towers.
Atkinson (2014, p. 7-8) notes that these visible memory apparatuses caused observers to comment
that “the sight of the reels moving back and forwards as the computer searched for stored pieces of
data gave the impression that the machine was ‘thinking,’” thus furthering the vision of a computer
as a “mechanical brain.”
Like Bush’s memex, the electronic brain’s memory differs in capacity and concept from the
memory that most social memory theorists of the 20
th
century discussed. If perhaps conveniently left
ambiguous, memory here was data storage, a necessary component of solving “abstruse” problems,
and evidence for these novel devices as sophisticated problem solvers, with influence in the social
world. Despite these differences, the enthusiastic public embrace of “memory” as a core function of
how novel computational devices operated, and the capacity of memory to order the world beyond
human faculty, suggests a linkage that is far from incidental. Indeed, it could have been another
way–why not call these functions, as Aiken insisted, “the storage of numbers”?
Novel machines from the memex to the electronic brain seemed to provide a means to order,
store, and recall information across time, replicating and extending human thinking. Memory, in
these visions, meant the faculty of knowledge recall––perhaps not so different from the “art of
memory” that Yates (1966) describes––just as it also became synonymized with computer data. At a

48
time when the public (or scientists in public, like Bush) were faced with feelings of information
abundance in the wake of the technological development of the second world war, machines were
seen as a means to process this abundance. Perhaps more importantly, they did so at a time when
the American public was grappling with the destructive power of technology in the wake of the
bombing of Hiroshima and Nagasaki, grappling with whether technology appeared within or outside
of human control (Wardrip Fruin, 2003). The vision of the memex and the electronic brain are
indeed visions of how humans and computational machines would interact, visions which had
memory at the center. The self-conscious emphasis on computers’ memory functions indeed put
forth a vision of novel technology that was more neutral, and less obviously destructive, than others
wrought by the military-industrial complex. That computers and other digital machines possessed
memory, could be used to extend human memory, and were ultimately better at remembering than
humans, indelibly linked these two entities together. Computational machines became unfailing (if
not entirely trustworthy) tools for the recall of the informational past.
Ubiquitous Computing in the Early Aughts
Fifty years after the memex was first conceptualized and anthropomorphized metaphors of
computer memory were in popular circulation, the image of what a computer was and what it was
used for had changed for the US public. The growth of digital personal computing in the 1980s and
1990s shifted the meaning of computers from large military devices in the 1940s (and the human
‘computers’ that operated them), to the offsite business devices used for time-sharing in the 1960s,
to smaller devices used in the office, the home, or at school. Because computers had shrunk to
‘desktop’ size, and were used to produce, organize, store, and navigate information by individuals,
technologists routinely drew parallels to Bush’s memex (Ross, 2003; Leslie, 2020). These associations
grew as the web gained prominence (Crovitz, 2012), and in the early to mid-aughts they seemed to

49
reach a fever pitch
3
as a spate of projects emerged that attempted to operationalize computers, the
web, and inexpensive digital storage toward the transformation of personal memory. For their
creators, these projects signaled a transformation in the construction of the self and its relationship
to the past. Known as ‘lifelogging,’ they included the Microsoft Research-based MyLifeBits (2004),
the US Defense Advanced Research Projects Agency’s (DARPA) project LifeLog (2003), and the
British Computer Society’s Memories for Life (2004).
Why did these earliest years of the new millennium produce such a rash of digital memory
projects? For one, computers and the web had become an intractable part of American life. Up until
the year 2000 and in the time immediately following the dot-com crash, it was up for debate whether
the internet would be a ‘fad’ or a persistent and lasting technology, transformational in nature in the
way some had said it would be (Sapnar Ankerson, 2018). At the same time, personal sharing using
computational devices and computer networks–whether “webblogs” or social networks associated
with Web 2.0 like SixApart, Friendster, MySpace–were rapidly gaining prominence. The
combination of these ‘social’ ideas, ubiquitous personal computing, growing access to the internet,
as well as rapidly expanding storage capabilities for rapidly decreasing costs (i.e. Moore’s Law) (Dix,

3
Memory and media seemed to be of interest in popular consciousness in general. This era saw numerous
memory-related movies released, all of which used media, in some form or another, to play with the
boundaries of remembering and forgetting: these included the psychological thriller Memento (2000), Final Cut
(2004) the cult hit Eternal Sunshine of the Spotless Mind (2004), the French arthouse psychodrama La Moustache
(2005), and the romantic comedy Fifty First Dates (2004). There are more: Jesse Kavadlo’s (2013) discussion of
memory in film cites thirty films released between 2000 and 2005 that involve amnestic characters. Memory
was on people’s mind. The reasons for this are not just computational: memory has long been a rich trope
through which to play with narrative structure; there was a profound focus on memory in the wake of
September 11
th
(ibid); and insights into memory from cognitive psychology and neuroscience may have
provided creative inspiration to these projects (Nungesser, 2009). [And then there’s the culture industry
(Adorno and Horkheimer, 1947).]

50
2002), prompted this series of experimental projects around digital technology and personal
memory.
I review these projects to show how imaginaries around computer-mediated memory
developed as computing became ubiquitous for personal purposes, analog media underwent
digitization, and the web embraced expressions of personal identity. Built on the early vision of
computers as reliable machines for the organization of the past and on the transmutation of memory
to data, these projects connect mid-century imaginaries around machines and memory to the
platformized visions of memory I discuss later on, thus establishing a basis for the types of memory
work that social media platforms embraced in the years that followed.
MyLifeBits
The most well-known of these digital memory projects was MyLifeBits, a project started by
the computer scientist Gordon Bell. Bell was famous in the computing industry for having been a
primary architect for Digital Equipment Corporation’s (DEC) computers. It was at Microsoft
Research that Bell made a play as a technological futurist, envisioning what life would be like given
imminent and unavoidable transformations in technology. Inspired by Bush in both prophetic
proclivity as well as in his interest of applying emerging technological systems for functions of
individual memory, MyLifeBits was Bell’s signature project.
The concept for MyLifeBits had been introduced in a 2001 Communications of the ACM article,
entitled A Personal Digital Store (Bell, 2001; Gemmell, Bell, Lueder, 2006a). In it, Bell and colleagues
sketched out a novel means to organize, search, annotate, and use personal content in one database
(Gemmell, Bell, Lueder, 2006a), as opposed to organizing disparate content in “data islands”
scattered on multiple devices and disconnected in non-compatible formats. In manifestos about the
project, Bell drew on the vision of Bush’s memex to explain the impetus behind the project, noting
that (Gemmell, Bell & Lueder, 2006a, p. 91):

51
Memex, with links and comments holding a central role, served as our blueprint for
MyLifeBits. Faced with folders full of documents, messages, phone calls, photos, and music
files, with inherent or potential metadata such as author, camera, comments, location, and
time, we needed a framework to hold and link all of these objects in the web-like and almost
arbitrary fashion that Bush described.

Bell was the architect of these schemes as well as the guinea pig, and in initial phases of
MyLifeBits the project consisted simply of digitizing all of Bell’s media: DVDs, photos, records, and
more for collection in this unified database (Gemmell, Bell & Lauder, 2006b). It then folded in the
capture of his data in the present: emails, readings, music explored, files opened, web pages visited,
phone calls conducted, television shows watched, radio shows listened (Van House & Churchill,
2008). A GPS system recorded his location. Bell also began using what was dubbed a ‘SenseCam,’ or
a camera worn like a necklace, that took and transmitted photos to the MyLifeBits repository every
30 seconds.
As Van Dijck has noted (2007), MyLifeBits, and digital memory projects like it at this time,
aimed to solve memory’s ‘fallibility’ through digital solutionism. That is, they conceptualized
memory as a “fixed object” that can be encapsulated through perpetual mediation. In doing so,
projects like MyLifeBits implicitly built on visions of computing as a reliable extender of memory,
just as they implicitly evoke the ‘imagined user’ at the heart of visions of the memex and the
electronic brain, of those ‘men of science.’ MyLifeBits also recalls the language, goals, and ethics of
transhumanism: a philosophy and a movement native to high-technology cultures that focuses on the
“power of technology to transcend the limitations of the human body as currently evolved” (Kelty,
2008, p. 86). Transhumanism has envisioned this transcendence in part through memory: whether
enhancing memory recall during the ‘pre-transcended life’ (Bostrom, 2003) or envisioning the
‘uploading’ of one’s identity (in large part through uploading memories [ibid]) to a silicon chip,
reloading it when the posthuman world has dawned (Kelty, 2008). While primarily thought of as a
fringe movement, Christopher Kelty has noted how transhumanism merely crystallizes perspectives

52
that are likewise at work in mainstream thought in more subtle ways–like the memory projects I
discuss in this section, or the memory functions of platforms I’ll discuss later on.
Memories for Life
Memories for Life was a “grand challenge” set forth by the British Computer Society in 2004
by early developer of the web and the organization’s then President, Dame Wendy Hall. Memories
for Life, like other Computer Society Grand Challenges, served as a node around which researchers
from multiple disciplines (though primarily computer science) could gather to catalyze research
around a particular topic. Like MyLifeBits, this challenge’s aim was to engage with the intersection
between computing, computer memory, and personal memory.
Memories for Life was less grandly prophetic than MyLifeBits, less hyperbolic about the
changes that might be wrought and the potential of those changes for society; the uniquely inflated
rhetoric of American techno-cultures, however, may account for this difference. It nonetheless
identified memory as primarily a means of recording, storage, and access, an expansion in recall
wrought by an expansion in the storage capabilities of computers for the average person. Despite
recognizing the significant hurdles in data maintenance over time, the challenge was conceptualized
around the possibility of perpetual and complete memory, or memory for life (O’Hara et al, 2006:
351):
Memory for life is a research problem, and a problem space—but what problem, and why
now? The use of electronic media for supporting human information storage and recall needs
defines an area. We have always had artificial aids to memory of course; the twenty-first
century twist is that suddenly we are presented with the possibility of memories for life.

Envisioning a gathering of research from computer science, information science, sociology,
neurobiology and psychology, Memories for Life nonetheless framed memory in a way that was
contiguous with the visions of computing from midcentury. It favored engineering-oriented and

53
positivist approaches that sat uneasily with social theory and qualitative approaches to memory
developed in disciplines outside of engineering (Van House & Churchill, 2008).
LifeLog
LifeLog was DARPA’s–the agency known for launching the Internet and for making
‘advances’ in military technology–own response to information proliferation and computer-mediated
memory. Inspired by Bush’s Memex, MyLifeBits, and Grand Computing Challenge Memories for
Life (DARPA, 2004), LifeLog envisioned computer-aided memory not so much as a means to
perfect recall for personal or scientific reasons, but rather as a means to aid the human ‘warfighter,’
at least according to the initial public plans. As an archived version of the LifeLog page on the
DARPA site reads (DARPA, 2004):
The LifeLog Program addresses a targeted and very difficult problem: how individuals might
capture and analyze their own experiences, preferences, and goals. The LifeLog capability
would provide an electronic diary to help the individual more accurately recall and use his or
her past experiences to be more effective in current or future tasks. […] The goal of the
LifeLog is to turn the notebook computers or personal digital assistants used today into
much more powerful tools for the warfighter.

While LifeLog promised to be a device in which the user would have total control, the public was
skeptical. A New York Times article criticized the project for its inherent potential to violate the
privacy of any individual who used it (Safire, 2003), and its creator’s contract was cancelled not long
after. Journalists have since referred to the project as a “failed military cyber-diary” (Axe, 2018).
Indeed, all the lifelogging projects I discuss here have ended, despite goals of perpetual
storage and perfected memory.
4
By May of 2017, the Memories for Life page redirected to the

4
Interest in lifelogging through ubiquitous capture, however, is not as bygone as we might think. The day
before writing this paragraph, Facebook introduced [on September 9] a feature that mimics much of the
lifelogging capabilities from over a decade prior–a pair of Ray-Ban sunglasses with photographic capabilities.
Known as “Ray Ban Stories,” Facebook emphasizes its use for “memories,” a word they synonymize with
data. The sunglasses integrate with “the new Facebook View app, so you can share your stories and memories
seamlessly with friends and social media followers,” their introductory post notes.

54
university host page, having not been updated since 2009. By 2016, it was reported that Bell had
abandoned MyLifeBits–its constant information was too difficult to sort, wasn’t of use, and, he
argued, had largely become redundant because of social networks and smartphones (Elgan, 2016).
Finally, DARPA’s LifeLog shut down on February 4, 2004–which also happened to be the day that
Facebook was launched by Mark Zuckerberg in his Harvard dorm room. This coincidence has led to
conspiracy theories about Facebook being the second and more palatable version of LifeLog, a US
government project for spying on its citizens (DARPA, 2020; “Why did DARPA,” 2019).
While these conspiracy theories are, like many, far-fetched, they nonetheless point toward the
alignment between lifelogging—the use of computational devices for the wholesale collecting of
personal information—and social media platforms’ use in recording, storing, and organizing media
about the individual. Sites that would be recognized as contemporary social platforms indeed
emerged in the same era as MyLifeBits, LifeLog, and Memories for Life: along with sunset sites like
Friendster and MySpace, which came online in 2002 and 2003, respectively, Facebook went live in
2004, Reddit in 2005 and Twitter in 2006. Following the bombastic promises of lifelogging for total
memory capture, ones aligned with the perhaps even more bombastic promises of transhumanism,
social media platforms positioned themselves around memory practice in more subtle ways. In what
follows, I show the ways that social media platforms–especially Facebook–embraced reflection by
integrating features for the capture, sharing, and persistent organization of moments from everyday
life.
Platforms Embrace Memory
Unlike lifelogging, social media platforms did not begin with the express desire to be used for
memory. If anything, they began from a desire for novelty. Friendster, after all, was rumored to have
been a pet project of its founder when he wanted to date new people, having exhausted the options
in his immediate social circle. MySpace based itself off Friendster’s design, its founders arguing that

55
its value-adds were greater self-expression and discovery of new music. Moreover, early research on
the social dynamics of these platforms observed the ways that users, especially young people, used
social media as a means to actively construct self-identity (boyd, 2014), not create a record that their
grandchildren could look back on at the end of their life.
At the same time, these platforms stored and organized personal user information so that
users could continue conversations with one another asynchronously and persistently represent
themselves to other users in the network. In doing so, platforms reproduced the mediated vision of
memory that had developed alongside the popularization of computing in the 20
th
century–one in
which memory was significantly linked to the storage of personal data on digital devices. What’s
more, the storage of personal media over time was crucial to the central social functions of these
platforms that internet researchers were identifying, including self-expression, identity construction,
and community. For instance, research examining social media use argued that the identity
construction taking place on platforms was based in the connection between past and present selves
(Salimkhan et al, 2010), as users either uploaded media from the past or accumulated media through
time on the platform. Practices like “throwback Thursday” or “flashback Friday,” popular on
Twitter, Facebook, and Instagram, attest to this practice, as users remediate content from the past in
the present. In forming community online, persistent information storage for reflection was also
important. As De Kosnik (2016) has noted, stored information acts as a record through which other
users can be socialized into particular online cultures. In offering personal data storage, platforms
implicitly built on the relationship between mediation, memory, and personal computational devices,
even if they positioned themselves (or were positioned by others, as I’ll discuss soon) as being
defined by liveness and novelty.
Storage-related features on platforms also referenced mnemonic media; in turn, mnemonic
media metaphors were sometimes used to explain these platforms to a broader public. By way of

56
example, a Guardian journalist writing about MySpace’s acquisition in 2006 noted: “If the blog has a
common ancestor with the diary, MySpace shares at least some of its DNA with the scrapbook”
(Lilley, 2008). In title alone, early aughts sites like LiveJournal had made the connection between
mnemonic journals and blogs, while early social network Classmates.com employed the memory-
inflected yearbook to structure its design. Photo hosting was also a major draw, as internet users
employed photo-specific sites like Flickr or uploaded images, from selfies to dorm parties, into
‘photo albums’ on Facebook, Friendster, or MySpace. Facebook early on introduced the automatic
saving of user profile pictures into an album, too, images that had previously been deleted when one
put up a new photo. What is more, sentiment that may have otherwise been embodied and
ephemeral, communicated in conversation, was now made textual and persistent through features
like ‘status updates’ that were core elements of sites like Friendster and Facebook. Platforms, in sum,
embraced the mnemonic role afforded to them through data storage.
As platforms aged, they built on their role in memory practice beyond offering persistent
storage for diverse media. They introduced organization schemes and features that helped in
weaving a narrative about the self that would be supported by the platform over time. Many of these
features and schemas emerged in the 2010/2011 era, at a time when US users had largely settled on
using Facebook as their profile-based social network. In turn, users had also largely acquiesced to
the idea of a platform that required users to be their ‘real self,’ forgoing the experimental subjectivity
of the early web for “known” and “persistent” online identities that aligned real life and online
personae (Deseriis, 2015, p. 168), with the use of a real name, affiliation to real institutions and
locations, and photos of one’s real self (for the most part). Facebook introduced features—like
previously discussed Timeline––that emphasized its role in telling a life story. These features
continue to be iterated upon today through the algorithmic resurfacing of events influenced by the
metrics (likes, comments, etc.) that content originally received (Jacobsen & Beer, 2021). They are

57
features that, like lifelogging before it, resonate with dreams of transhumanism: where one’s ‘whole
life’ could be represented through mediated uploads.
The notion that one’s ‘whole life’ would be represented on the page, beginning to end, began
to become a reality as social media platforms encountered user death. Companies created
sociotechnical structures to respond to this memory-intensive moment.
5
Recognizing that the media
uploaded there often continued to have importance to a deceased user’s friends and family,
platforms produced avenues through which users could notify the platform of a user’s death and
freeze a user’s account, allowing it to be used as a site of remembrance. In 2009, Facebook for
instance debuted a “memorial” setting, which changes a profile from an active one with the person’s
name–say, John Doe–to one that instead reads “Remembering John Doe.”
In addition to Facebook’s Timeline, third party apps promised to organize the accumulated
online past in an engaging way. Timehop, founded in 2011 and continuing to exist today, was the
most prominent example of these apps. Aggregating posts from multiple social platforms–including
Facebook, Instagram, Twitter, and elsewhere–it provides an individual user with updates about what
they posted on that day years ago–perhaps one year ago, perhaps ten. Memolane, also launched in
2011, likewise arranged data from multiple sources into a personal visual timeline. The founder
called it “a modern-day scrapbook that writes itself” (Wortham, 2011). Projects like Intel’s popular
“Museum of Me,” introduced in (when else) 2011, mined Facebook content to create a museum like
experience of content that had been uploaded to the platform. Facebook itself has integrated many
of these features into its own proprietary structure, such as TimeHop mimicker “On this Day”–
introduced in 2015 (Meta, 2016)–and algorithmically generated ‘movies’ created from highlights
chosen at the end of the year.

5
Silicon Valley-based startups that promised to manage digital assets after death also began to emerge in this
era (see Kneese, 2018).

58
Whether through flashback apps, photo albums, memorial sites, or simply persistent user
profiles, social platforms built on the alignment between computational media storage and user
memory. This alignment resonates with the intensification of media use in the organization of the
past, as evinced by memory studies scholarship, just as it reflects the vision put forth during the
popular debut of computing in the mid 20
th
century, wherein the meaning of digital data storage–
computer memory–blurred with memory as the social process of reflection.
Memory: A Typifying Feature of Platforms
What this brief timeline posits is that platforms have been defined by data storage over time
and its relationship to memory practice. They are defined by it in the sense that data storage is a
fundamental aspect of what gives meaning to platforms. That platforms are defined by data storage may
not be a widely shared view. It has been suggested that while platforms have elements of storage
that are useful for memory, their core offering is live communication and informational novelty. As
Bernie Hogan and Anabel Quan-Haase write in “Persistence and Change in Social Media,” a special
issue and review of major themes of social media scholarship prior to 2010:
In theory, data submitted to social media can be persistent. In reality, most social media
practice is oriented toward the shared present. Users constantly update data and the traces of
the data vanish in the archives. Social media focus on the unfolding in the here and now of
cultural, environmental, and political events and people’s lives.

Overt reflection on stored media may indeed be rare: scrolling through content to see what you were
doing seven years ago is likely not an everyday practice. At the same time, this lopsided rhythm is the
cultural norm for memory-invoking media like photo collections, diaries, family letters: artifacts
looked upon rarely and usually under particular circumstances. To analogize further: just because a
person tunes in to the evening news every night does not mean that their home videos are made
incidental. Liveness, that is, does not preclude memory as being a typifying feature of platforms.
However little used years-old information may be in comparison with the constant updates that are

59
pushed to the top of the feed, social media platforms would be unrecognizable without the ability to
organize and navigate media from the past, looking more like early aughts chat services AOL Instant
Messenger or ICQ that enabled conversation but deleted it from public view after a chat session was
closed.
More accurately, most people don’t regularly scroll through their content to see what they
were doing seven years ago because memory features have been so seamlessly integrated into
platform design; that is, the past is also pushed to the top of the feed. Mediation of the past is so
ingrained, so fundamental to social media platform experience that it is rendered invisible. Lee
Humphreys (2020) has recently argued that platforms’ memory features–like Facebook Birthdays or
On This Date–are largely overlooked because they are “routine and less momentous than political or
economic affairs that dominate the public sphere” and they continue analogue features, which
makes it less visible than other ‘disruptive’ forces of digital society.
6
As she notes, and as I concur:
These patterns can be overlooked because of their habituation. Yet, these everyday practices
and structures also reveal deeply embedded sociocultural values and, in this case, the
economic and structural influences of media that shape the production of media content as
well as mediated time.

6
If platforms were appealing to wide swaths of users in part because of their ability to construct identity
through memory, then what do we make of ephemeral messaging platforms like Snapchat? That app’s value-
add was in offering ephemeral messaging, and it inspired the subsequent embrace of temporary posts on
platforms like Instagram and Facebook, including features like ‘Stories’ and ‘Facebook Live.’ But it is also a
mistake to think of these ‘ephemeral’ messaging sites as anti-memory. A study by McRoberts, Ma, Hall and
Yarosh (2017) showed how Snapchat users employed the platform’s 24 hour ‘Story’ feature as a buffer period
to decide whether they would like to save this otherwise ephemeral content. Multiple interviewees noted that
they enjoyed reflecting on their recently posted Stories, and occasionally saved snaps, sometimes to share on
other (more permanent) social media. In what is an extremely limited sample, I have asked students in
courses I have taught what tools they use to organize and reflect on their past. Counterintuitively, the
majority cited Snapchat as a primary means of organizing their personal memories online. They modify
settings such that posts they send are saved, a private album of what would otherwise be thought of as time-
limited media. Snapchat indeed added the “Memories” feature in 2016 in response to users wanting to be able
to save media that they had posted.

60
Crucially, these reflection and storage features are not only placed there because they are useful tools
for existing users. Persistent engagement with a platform gives impetus for users not to leave: why
delete your profile when you’ve got a decade of content stored there? This is to say that platform
definition––and platform business plans––are intimately tied to a human relationship to the past.
The storage of meaningful media, which evokes moments from a personal past, is a significant play
for persistence and enduring usage. In other words, it’s a play to become indispensable. To put it
differently, it’s a play toward becoming infrastructural.
Digital Memory Infrastructure
Embedded, intractable and invisibilized means of organizing, storing, and serving up the
past, platforms are a crucial part of what I argue are digital memory infrastructure: those computational
systems that underly, support, and shape modern memory practice. Social media platforms exhibit
characteristics that STS scholars commonly identity as infrastructural, and they exhibit these
characteristics for purposes of memory, for reflecting on and constructing the past. For one,
infrastructure has been described as those underlying systems which support modern societies
(Carse, 2016), and platforms are indeed underlying and ever more embedded systems in the media-
intensive memory practices of today: from mourning (Wagner, 2018) to reminiscing (Jacobsen &
Beer, 2021); from serving as a persistent record for community socialization (De Kosnik, 2016; Villa
Nicholas, 2019) to shaping the collective memory of a public event (Smit et al, 2017a; 2017b); just as
they help users remember friends’ birthdays (Humphreys, 2020).
Platforms, as I’ve argued in this chapter, were not founded around memory, yet these
companies both embraced user memory practice as it was tied to media hosting just as they pursued
a more pronounced role in mediating users’ relationship to the personal past over time. In this
process, platforms’ role in memory practice has come to reproduce characteristics of infrastructure
that have been identified by STS scholars. For instance, Susan Leigh Star’s (1999) well-known tenets

61
of infrastructure include relational categories like embeddedness, transparency, links with
conventions of practice and being learned as part of membership, among others. When put in
conversation with memory, platforms indeed satisfy these principles, as they are, respectively, sunk
into social structures (e.g. remembering communities like the family or community) and other
technological arrangements (the greater computational ecosystem); they invisibly support memory
tasks, as argued in the previous section; they interface with existing memory conventions (e.g. media
like photographs were primary vehicles for reflection before platforms existed); and new users
develop naturalized familiarity with infrastructural conventions as they are socialized into that
platform’s culture (e.g. that Thursdays are a time to remediate photos for ‘Throwback Thursday’).
Yet a tension arises between platforms’ role in memory practice and definitions of
infrastructure, in that infrastructures are traditionally marked by permanence and stability; as Paul
Edwards has stated, infrastructures provide “the stable foundation of modern social worlds” (2003,
p. 186). Can platforms really be digital memory infrastructure if, as I argued in the Introduction, the
history of these sites and services is marked by data loss, shutdown, and failure–marked by
instability?
On the one hand, extant platforms—especially Facebook and its various sub-platforms—
have consolidated market power and presented users with high switching costs (Tucker, 2019), both
in relation to memory but also in relation to the other basic functions that the platform supports
(e.g. Plantin et al, 2018). In that sense, they have stabilized more than other platforms I focus on in
this dissertation. Sites like Facebook, Twitter and Reddit, moreover, undergo routine maintenance
that counteracts the otherwise volatile nature of data storage. As Star and Bowker have argued
(2010), information infrastructures that rely on data storage require ongoing work to remain
‘permanent’ and stable:

62
The bottom line is that no storage medium is permanent, […] so that our emergent
information infrastructure will require a continued maintenance effort to keep data
accessible and usable as it passes from one storage medium to another and is analysed by
one generation of database technology to the next.

At the same time, because infrastructures are not absolute entities but exist relationally, they are
inflected with cultural expectations, a notion I develop more fully through the concept of
‘infrastructural expectations’ in Chapter 5. The ‘electronic brain,’ the memex, and lifelogging asserted
a vision of computational memory that was permanent and stable, juxtaposed as they were against
‘fallible’ human memory. I suggest that these cultural visions of computational memory were later
embodied by platforms, such that the expectations users bring to these sites are expectations of
stability––whether this is accurate or not––but nonetheless further ensconcing their status as digital
memory infrastructure.
As I have shown in this chapter, social media platforms have built on the cultural alignment
between computational media storage and user memory. In doing so, they have been transformed
into digital memory infrastructure. As I noted in the beginning of this chapter, identifying platforms
as digital memory infrastructure is a categorizing moment. It recognizes that these systems shape the
construction of the past for individuals and for communities, and to a broader extent, for societies.
This categorization informs a basic premise of this project, that is, that platforms as digital memory
infrastructure continue to act as structuring forces for the past even as they decline and sunset.
These systems–embedded, relational, linked to conventions of practice and interfacing with existing
social structures–don’t simply disappear, severing all material and social links when they are
unplugged. Instead, in their transition to ruin, they continue to shape a relationship to the past. How
these infrastructures ‘live on’ through their modes of shutdown and erasure, how they are inflected
with values, and how they continue to shape the production of the past, are addressed in subsequent
chapters as I turn to the empirical component of this study: the act of sunsetting.

63
Chapter 2

Vicissitudes of Value: The Commercial Logics Behind Digital Loss

A roller coaster, is the way it was put to me. You’re going up like this, and you just start
going over the edge a little bit, then all of a sudden it gets momentum, and then it just shoots
[down].

– Louis, database administrator at MySpace

The hockey stick graph is emblematic of Silicon Valley social media platforms (Figure 4)–or
at least it represents what most social media companies hope happens to them. A short period of
limited growth, known as the blade years, an inflection point, and then a limitless, exponential
uptick. The hockey stick shows a company without decline, a future that goes up, and up, and away.
What the graph specifically models vary: while the bottom axis is always time, the upward swing
could be the number of users, new user sign ups or total user time, it could be its percentage of
market adoption, or it could be the company’s revenue or profit. Regardless of what it tracks, what
the image of the hockey stick communicates is the same: runaway growth in a platform’s value.
Value is a broad concept, signifying at its most basic the relative worth, usefulness, or
importance of something (Value, n., n.d.). It is the concept this chapter hinges around, focusing
specifically on both cultural value and economic value as they are articulated by platform
organizations and their employees. Cultural value and economic value are themselves expansive and
internally heterogeneous, but in this case are useful heuristics for understanding the changing forms
of valuation that platforms are subject to over time. Cultural value, which I address in more detail
later on, can be understood as the relative worth of something within a particular social milieu, while
economic value can be understood as the relative worth of something in so far as it can be
converted to money, a definition that draws on related characterizations of ‘economic capital’
(Bourdieu, 1986; cited in Sadowski, 2019). Importantly, the economic value of a given platform can
signify not only of what could be called realized economic value (e.g. actual revenue by selling

64
advertising space), but also speculative value, communicated through a company’s ‘valuation,’ or a
variable number that roughly approximates what the company would be worth if it was sold.
Valuation can be based on factors as varied as “a team’s expertise, product, assets, business model,
total addressable market, competitor performance, market opportunity, goodwill, and more,” as
Silicon Valley finance company Brex has phrased it (Brex, 2021). This value is speculative in two
senses: because it conjectures about realizable economic worth based on variable evidence, just as it
represents value that is conjectured about because of interested parties’ pursuit of economic gain
with the prospect of monetary risk. In other words, speculative economic value represents different
stakeholders’ beliefs about a company’s monetary potential.
Indeed, the value communicated by the hockey stick graph is largely economic, and
speculatively economic at that. As one industry think-piece on fostering company growth described,
startups seeking “angel and venture-capital funding are conditioned to project a ‘hockey stick’
revenue curve” (Deutsch, 2017) a promise if not a likely reality of the economic bounty to come,
and therefore an invitation for investors to get in on the ground floor before that upward
momentum takes off. For social media platforms, the projection of economic bounty communicated
by the hockey stick graph does not necessarily mean that revenue or profit are directly measured:
rather, for these sites, user metrics can also represent the promise of economic value. A business
manager for GeoCities summed up this ethos when he noted that their early goal was to “get as big
as you can, as fast as you can. Then any idiot will be able to figure out how to monetize all that
traffic.”

65

Figure 4. Hockey stick graph (Biery, 2016)

When compared to the innumerable platforms that have come and gone without much
fanfare, the sites I focus on in this dissertation tended to be the lucky ones, at one point in time
experiencing this much ballyhooed hockey stick growth and the resultant belief that their site had
economic and cultural value. From GeoCities to Friendster, MySpace to Vine, along with less well-
known platforms like Bolt or Friends Reunited, the employees I spoke to often got to experience the
heady moments that followed the graph’s inflection point. They had the chance to revel in the idea
that their company was one of the ones that would succeed, that the company they were working for
held economic value, and indeed that the company was offering something of value to its users. As a
founder of mobile platform Couple recounted, there are few problems when metrics track upward:
“There is this saying: growth fixes everything.” Indeed, employees who worked for these platforms
during their early years were often astonished at the magnitude of the hockey stick’s upswing. “We
would sit there and just look at these charts of our growth,” one Friendster staff member

66
remembered, just as another recounted how, “there were days we sat there and were like, we've
never seen anything like the growth that we're experiencing anywhere in our careers.”
Social media platforms famously benefit from “network effects,” or the phenomenon that
the value of the platform–in this case, its value to its users–increases the more users it acquires
(Stobierski, 2020). To put it more simply, people want to join a social media platform that has many
users rather than few. People want to join a network that their friends are on. Hockey stick growth
happens as these network effects kick in, especially because the platform typically comes to
dominate its market domain (Hagiu & Rothman, 2016). But the economic value of a platform does
not grow simply because more people have signed up. Rather, the fact that these users generate and
consume media using the platform increases its value to users–after all, just as one would prefer to
join a platform that has more users, the platform is only interesting so long as the users are doing
interesting things on it, so long as users have something, and someone, to engage with. The
generation and consumption of media by users in turn increases the platform’s economic value, as
user engagement with content provides ample opportunity to show advertising during a browsing
session, or to mine user engagement data for saleable insights. What I want to point out in tracing
this growth calculus is this: the hosting of user generated content–a crucial component of digital
memory infrastructure–helps generate a platform’s economic value.
Up until a point. It is a point that is not represented by the hockey stick of growth, a point at
which the graph line over time starts to flatten, to wobble, to tip, to decline. It is a point when user
engagement and user numbers slow and the economic value of a platform decreases. And it is a
point at which user-generated content that was previously made and then stored on platform
infrastructure fails to generate the economic returns it once did. In fact, as I show in this chapter,
user-made media hosted by the platform can move from an asset to a burden from the perspective
of the platform company. Picking up where graphs of hockey stick of growth leave off, this chapter

67
examines this process wherein platform components are emptied of their once perceived value, and
notions of what has value, and where value resides on a platform, start to shift.
Focusing on phases of platforms where user numbers, user engagement, and economic value
declined, this chapter draws on two cases to analyze vicissitudes of value during these moments.
One is Friendster, the dating site qua platonic social network founded in San Francisco in 2002,
where users could connect with one another while viewing their degrees of separation from other
users in the social graph. Much like Facebook today, Friendster users were encouraged to use their
real names and images on their personal profiles, and could send messages, post photos, and
eventually see others’ activity within the network via a centralized newsfeed. The other is MySpace,
the Los Angeles-based and comparatively customizable platform commonly cited as overtaking
Friendster as the main profile-based social network in the United States not long after its launch in
2003. I situate these cases within literature on the assignation and transformation of commodities’
value, focusing on texts that address commodities’ transformation into valueless entities, into ruin or
waste. Ultimately, by illustrating how digital loss is shaped by shifting valuation during periods when
a platform is thought of as in decline, I attest to the presence of–and the importance of articulating–
the commercial underpinnings of deletion through the concept of commercial digital loss.
Rubbish Theory: On Salvaging Value
“What can we salvage?”
That’s a question MySpace employees broached as they worked to sunset the classic
MySpace platform in 2013 and bring it into its next era under Specific Media, their new owners.
Specific Media has purchased the beleaguered MySpace two years prior to that, for $35 million
dollars–more than $500 million dollars less than what it had been purchased for when News Corp
acquired it from Intermix Media in 2005. Michael, a content producer who worked there during that

68
transition, described how the organization asked this question in the context of what information
from the old platform should be carried forward on to the new one they were building.
Salvage was an apt term. Salvage traditionally signified goods or cargo saved from a sinking
ship (salvage, n., n.d.), a metaphor that employees had indeed used to describe platforms as they
declined. It was also apt in that salvage in the 20
th
century came to mean the act of saving and
collecting waste material, typically for recycling. It conjures salvage yards, open-air lots where
decommissioned vehicles are brought, sold for usable parts or dismantled for their scrap, which is
then recycled and reformed and put forth toward something new. Salvage yards are spaces where a
commodity comes to be squeezed of its last remaining value, lest it otherwise be left as rubbish. The
decline of a platform and the decision to sunset it often resemble processes like these: wherein
organizations assess diminishing value of their assets, deciding which components of the
sociotechnical platform system are rubbish and what could be salvaged, sold for scrap.
Beyond platforms, anthropologists and other social scientists argue that rubbish and
salvage–things emptied of value or with diminished value–play important social roles. Michael
Thompson (1979), in his book Rubbish Theory, proposes that waste and rubbish are crucial to social
systems of value, despite the tendency for rubbish to be rendered invisible: hidden away or
disregarded because the object in question possesses no value or even negative value. Thompson
argues that visible objects tend to fall into two categories. The first are objects that are transient, for
which value diminishes over time and which has a circumscribed lifespan–like a smartphone which
eventually will obsolesce, perhaps one day ending up in an electronics junk yard (see Gabrys, 2011).
The second is durable, or something that maintains value over time and has a potentially infinite
lifespan, like an antique piece of furniture or a painting from a well-known artist. In between these
categories there is a “region of flexibility,” the anthropologist notes, in which those with social

69
power can enforce what is transient and what is durable, typically ascribing their own objects with
durability (Thompson, 1979, p. 9).
7

Thompson argues that rubbish falls outside of these two categories, instead offering a third
category that is seldom considered given rubbish objects’ relative invisibility. Rubbish, he argues, is
an important channel through which objects can move from transient to durable. It is a conduit
through which the qualities associated with an object can be altered, creating value in this process.
He provides the example of a Stevensgraph, or small woven silk pictures produced in the Victorian
era in Europe, typically depicting horse riding and cricket matches and other leisure activities of the
gentry. Sold for a shilling at the time, they were accessible and mass reproducible art. But the
ubiquitous nature of the art and changing tastes of the populace caused the Stevensgraph to stop
being produced and stop being sold at all–antique dealers would not purchase them as they were
virtually unsellable kitsch. They were rubbish. Through the choices of individual collectors, however,
who identified some special aesthetic quality in these items, Stevensgraphs in the 1960s became hot
commodities–their value skyrocketing, their status now moving from rubbish to durable. Thus,
Stevensgraphs went from relatively transient art–something that was in fashion but fell out of it; to
rubbish, with no value; to durable, completely and rapidly. (He even accompanies this by a graph of
a Stevensgraph’s monetary value over time: it looks like a hockey stick [p. 17]) As Thompson argues,
this sudden rise in value also tracked with the art’s changing social association, from the feminine
realm of the home to the masculine realm of learned institutions, as these works began being
included in museum shows and art books.
What this extended explanation of rubbish theory and Stevensgraphs amounts to is the
notion that rubbish plays an important social role, and tracking valuation processes toward states of

7
Thompson’s (1979) note on the differential nature of classification recalls Pierre Bourdieu’s work on cultural capital, in
that higher social classes determine what commodities and behaviors are associated with power and class.

70
waste, salvage, and rubbish can illuminate not only the objects in question but also the social
structure shaping their valuation. This work also shows that the qualities and value of an object are
not inherent in the objects themselves–not materially determined–but are placed on them socially.
(It is an idea also expressed through the saying “One man’s trash is another man’s treasure.”) This
perspective tracks with the argument of fellow anthropologist Arjun Appadurai (1986), who has
argued that objects’ value can shift, moving into and out of the commodity state–the state of having
economic value–over time through social exchange. And while Appadurai’s methodological
consensus in his analysis leads him to propose that analysts “follow the things” to understand the
social construction of value (p. 5), Thompson notes that to “study the social control of value, we
must study rubbish” (1979, p. 19).
The idea that waste is important to the study of value may not sound as daring as it did when
Rubbish Theory was published in 1979. Indeed, the lens of “broken-world thinking” (Jackson, 2014)
discussed in the Introduction resonates significantly with Thompson’s entreaty to look toward what
is labelled rubbish to see otherwise obfuscated social relations. Scholarship explicitly focusing on
waste and valuation, especially in anthropology, have further extended this theory, for instance,
arguing that value cannot exist without waste (Greeson, Laster & Pyyhtinen, 2020), or that waste
helps reveal even more complex transformations in value than Thompson observed with his own
hockey stick graph, with this relationship instead non-linear (Greeson, 2020).
8
Moreover, scholarship
addressing waste since Thompson’s publication has noted how it does more than just complicate a
transient/durable binary of value, but rather signals a broader and conflicting range of value that
Catherine Alexander and Andrew Sanchez have called “indeterminacy” (2019), in which objects can

8
See also Alexander & Reno, 2012; Hawkins & Muecke, 2003.

71
supersede the categorizing binary of valuable/rubbish, in so far as they are unrecognized by
dominant systems of classification.
While theorizations of waste have most prominently been taken up in anthropology, there
have also been useful interventions made in the study of cultural heritage and, to a more limited
degree, in the study of memory. Conceptions of value and conceptions of memory play well
together: this is especially because, as cultural memory scholars have shown, ideas and affective
orientations to the past are socially placed on to cultural objects, just as work like Thompson’s and
Appadurai’s have shown how qualities are socially placed onto objects.
9
Moreover, cultural
historiography–especially writing about groups at the social margins who were not considered
worthy of sustained archival attention–is often made through traces once considered transient or
rubbish (Shiozaki, 2021). Memory theorist Aleida Assmann (1996), for instance, has analyzed
contemporary literature to comment on the resonance between trash and trace, pieces of trash
eventually moving into a valued state as they come to represent “authentic traces of the past in a
mass media culture” (1996, p. 132).
Rubbish theory and its antecedents are useful in framing the valuation that occurred in
processes of platform decline and sunsetting, revealing the shifting value of platform components
after the hockey stick graph no longer accurately describes a platform’s trajectory of value. Rubbish
in memory and historiography establish a foundation for thinking through how trash can become
valued traces through their representation of the past, and through which to read the past. At the
same time, rubbish theory has been critiqued because it does not account for the practices that shape
an objects movement between categories of transient, durable, and rubbish (Fisher & Smiley, 2016;
Parsons, 2007). Understanding these processes and practices of valuation, and the movement of

9
One might also think about the centuries-old phrase: “the trash heap of history” (Safire, 1983).

72
platform components from valued objects to burdensome waste, are especially important, I argue,
because they help reveal the commercial logics behind content loss and digital memory
infrastructure.
The Breakdown and Decay of Infrastructure
I make a brief return to salvage and ships before directing course to platforms. A common
subject to studies of waste and studies of infrastructure has been the act of ship breaking. Ship
breaking refers to the act of dismantling large ships after a vessel is deemed unusable, whether
because of material degradation or inefficiency. Ships are deconstructed, stripped to isolate valuable
materials or still usable parts, then their parts recycled. Jackson (2014) uses ship breaking as the basis
for his development of “broken world thinking,” a perspective typically applied to the maintenance
and repair of infrastructures, just as waste theorists have employed ship breaking as an example of
how commodities move through states of valuation (Crang et al, 2012). The salvage act of ship-
breaking has provided a massively material example through which to examine processes of change,
whether sociotechnical dismantling or socioeconomic valuation, an object-based overlap that shows
the potential for productive dialogues between studies of sociotechnical dismantling from science
and technology studies and studies of socioeconomic valuation from waste studies. Inspired by
salvage and ships, I in turn combine these perspectives to bring them to platforms as memory
infrastructure, providing a means of concurrently addressing the dismantling of digital memory
infrastructure while also addressing changing valuations of platform components. Consequently, I
will discuss in more detail how STS approaches to infrastructure (within STS, anthropology, and
related disciplines) have approached infrastructural dismantling.
The decay and maintenance of infrastructures have been a growing scholarly focus as well as
an issue of public interest. In the United States, decaying infrastructure–especially for
transportation–has attracted the attention of politicians and a public concerned about ‘crumbling’

73
materiality of roads, bridges, train tracks, pipes and power plants.
10
In 2021, US Congress passed the
Infrastructure Investment and Jobs Act (IIJA), rallying massive capital investment to counteract
material degradation and “modernize” US infrastructure (McBride and Siripurapu, 2021). A major
reason behind the legislative push and the IIJA’s reasonably bipartisan support was so that the US
could remain “competitive” in a global marketplace (e.g. “President Biden's Bipartisan Infrastructure
Law,” 2021; Tomer et al, 2021), an anxiety that has likewise been at the root of US discourse and
investment in “innovation” (Vinsel, 2014a; 2014b). What this instance of infrastructural decay and
its social context illustrate is how infrastructural discourse and its social roles are significantly tied to
that infrastructure’s maintenance and processes of decay, even if the construction of new
infrastructure is also implicated in this discourse.
The same could be said for classic texts on infrastructure in science and technology studies.
Langdon Winner (1980) famously argues for understanding technology as embodying “politics,” or
specific forms of power and authority, because of its specific material configurations. At the heart of
this argument is the notion that these artifacts embody politics because of their modes of
persistence, decay, and required maintenance over time. As he writes (Winner, 1980, p. 127)
“societies choose structures for technologies that influence how people are going to work,
communicate, travel, consume, and so forth over a very long time.” Over this long period of time,
technologies continue to embody specific forms of authority through their required maintenance:
“to say that some technologies are inherently political is to say that certain widely accepted reasons
of practical necessity–especially the need to maintain crucial technological systems as smoothly working entities
[emphasis added]–have tended to eclipse other sorts of moral and political reasoning” (p. 133). More

10
While I use the word “decay” as a node around which to discuss infrastructural breakdown across public
and scholarly discourse, “crumbling” is what is often used by politicians to describe an urgent need for
investment in particular infrastructures in the United States (e.g. “The President promised to work across the
aisle to deliver results and rebuild our crumbling infrastructure” [“President Biden's Bipartisan Infrastructure
Law,” 2021]).

74
recent texts, especially those adopting STS approaches within anthropology, have argued that the
material decay, maintenance, and repair of infrastructures are central to the production of social
difference (Ramakrishnan, O’Reilly & Budds, 2021); in one example, analysts have shown how the
relations surrounding decaying water infrastructure in Rajasthan, India, helped reinforce caste and
gendered hierarchies (O’Reilly, 2006; O’Reilly and Dhanju, 2014). What texts like these–that position
modes of material decay as integral to the power of technological systems–prompt for the analysis of
digital memory infrastructure is the injunction to look to maintenance processes that sustain these
infrastructures to understand their embodied politics.
Yet by analyzing cases that are massively material in nature–roads, bridges, nuclear energy
plants, ship breaking, and such–these analyses rarely have trained their eye on breakdown in the
context of digital infrastructures, whose material visibility tends to differ (Jackson, 2016). This
visibility is inevitably technically as well as socially determined. Visibility is always dependent on the
perspective of the observer, as analyses of maintenance and repair have shown (e.g. Orr, 1996; Star,
1999). Moreover, perceptions of digital media as dematerialized help position even the massively
material components of digital systems, like undersea cables or data centers, as less visible when
compared to other infrastructures (Blanchette, 2011; Hogan, 2015; Starosielski, 2015). That is, even
in public discussion of these systems, attention tends to be diverted away from material
considerations, as dominant perspectives continue to understand them as frictionless and fluid
entities (Starosielski, 2015). As such, there has been little inquiry into the breakdown of digital
systems and especially digital knowledge infrastructures, or those “robust networks of people,
artifacts, and institutions that generate, share, and maintain specific knowledge about the human and
natural worlds” which rely on digital technologies (Edwards, 2010; Edwards et al, 2012).
While I return to material concerns explicitly in Chapter 4, as I examine the act of
dismantling technical components of social media platforms, in this chapter I bring together analyses

75
of the infrastructural decay of digital systems with discussions of shifting valuation. Though texts
like Winner’s (1980) position modes of breakdown as integral to the power of technological systems,
these analyses are less concerned with how processes of valuation are implicated in this process.
One hypothesis for this disconnection is that much of the massively material infrastructure
discussed in analyses of breakdown is that of government-funded entities, where forms of corporate
valuation may have less obvious import. Studies of infrastructural decay and studies of valuation
through waste, however, complement one another: sometimes trained on the same subjects, studies
of valuation highlight commercial logics underpinning infrastructural breakdown, while studies of
infrastructural breakdown bring a focus on practices that may otherwise be lacking from studies of
valuation. When it comes to analyzing platforms as infrastructure (e.g. Plantin, Lagoze & Edwards,
2018), however, these are more than complementary lenses. Rather, they are inseparable
components, as the breakdown of these systems is shaped by transformations in economic and
social value. It is this relationship that is parsed in this chapter, as I describe the landscape of
valuation during decline as well the components that are understood to retain or be emptied of value
during a platform’s process of decay, showing how these vicissitudes of value are not inherent to
some fixed technological lifecycle, but instead are produced by their social and economic contexts.
Indeed, along with showing the commercial logics that structure valuation during sunsetting, this
chapter reveals how forms of socioeconomic, racialized and geopolitical power are at work in
platform organizations’ assessments of what and who possess value as a platform goes into decline.
Friendster: For Whom Not all Hockey Sticks are Equal
11

Friendster is remembered within Silicon Valley circles as a progenitor of social networking
and as a cautionary tale of remarkable failure. Founded in 2002 by entrepreneur Jonathan Abrams,

11
Portions of this section have been published in Internet Histories: “Why does a platform die? Diagnosing
platform death at Friendster’s end” (Corry, 2021).

76
the site enjoyed skyrocketing users in its early years–three million users by early 2003 (Ulunma,
2020)–along with seemingly bottomless enthusiasm, as investors from technology’s most sought-
after venture capital firms jockeyed for their share in its auspicious rise. First built for those seeking
relationships, Friendster operated as both a dating site and platonic social network where users could
connect with one another while viewing their degrees of separation from other users in the social
graph. Much like Facebook today, Friendster users were encouraged to use their real names and
images on their personal profiles, and could send messages, post photos, and eventually see others’
activity within the network via a centralized newsfeed. While built around users in the United States,
Friendster had also become popular in Southeast Asia, including the Philippines, Malaysia,
Indonesia, and Singapore, where its user base rapidly exceeded that in the United States within the
site’s earliest years.
For all its success, however, Friendster was quickly plagued by slow load times, managerial
mishap and ascendant competitors as it tracked upward on its proverbial hockey stick of growth. In
the mid-2000s, US-based use tapered off as Friendster users switched to MySpace and Facebook,
and the company underwent numerous strategic changes while attempting to live up to early
expectations. According to insiders, the company tried to ignore its expanding userbase in Southeast
Asia, who appeared ‘unmonetizable’ to Friendster’s investors. By 2006, Friendster was considered an
also-ran by most US-based social network users and technologists. Pieces like the New York Times’
“The Fall of Friendster” (Rivlin, 2006) dissected the salacious details of the platform’s demise, while
its story was immortalized in a Harvard Business School case study (Piskorski & Knoop, 2006). In
the years that followed, the platform would be sold to Malaysian payments company Money Online
(MOL) in 2011, turned into a gaming site, and ultimately dissolved in 2018.
The 13 employees I spoke to from Friendster hailed from nearly all eras of the site’s tenure: I
talked to six employees there during its ascendance in 2002-2004, four employees during its decline

77
in US engagement and continued use in Southeast Asia in 2005-2007, and three employees from its
pivot to focus full-time on users in Asia and its attempt to find a buyer during 2008-2011. I did not
speak with any employees who worked there after its acquisition by MOL, during its transition to a
gaming site and eventual final closure in 2018. While not speaking to any employees from this era
limits my ability to discuss the final stages of sunsetting, and indeed how valuation may have differed
from the perspective of those situated in countries where Friendster remained popular, my
interviews do present the opportunity to understand vicissitudes of value as the “relative worth” of
the platform declined, at least as it was perceived by the Friendster organization and its employees.
Big, Indeterminate Value
Friendster’s story of valuation begins when the site’s value was high: high in number of
users, number of dollars of venture capital funding, number of minutes spent on the site, number of
articles buzzing about its growth. Certainly, this value indicated economic promise: by March of
2004, blue-chip Silicon Valley investors Kleiner, Perkins, Caufield & Byers (KPCB) and Benchmark
Capital had put $13 million dollars in the company, with a valuation set at $53 million (then large for
a company with only 12 employees and no realized profit) (Heilemann, 2003). By then, Friendster
had also fielded and rejected a $30 million dollar acquisition offer by Google (ibid).
At the same time, however, Friendster employees felt that the site’s value was indeterminate,
in the sense that the value it provided its users, the value of its enormous traffic, and the value of
fledgling social networking––what even constituted value for the site––was murky. Indeed, as noted
earlier, Alexander and Sanchez (2019) have argued that there exists a range of “indeterminacy” in
processes of valuation because some artifacts may fall outside of dominant systems of classification
(2019). While the Silicon Valley tech industry represents nothing if not a dominant system,
employees nonetheless engaged in wayfinding around the value of the network. What worth did the
company have for its users–or, as early Friendster employee Jerome noted his modus operandi was,

78
“Are we providing value to the customers?” And what worth–economic and cultural–did it
represent for its employees, and its investors?
My interview with Peter signaled this indeterminacy. Peter was an employee for Friendster in
these early years, a technical architect brought on in 2004 to handle the influx of users that routinely
crashed the site’s servers. He described this indeterminacy in value, but also signaled an awareness
that if the organization did not articulate what the platform’s value was––what its features and its
massive traffic meant socially, and economically, and in the context of Silicon Valley competitors––
then Friendster would soon slide off its hockey stick of growth:
My memory is that we knew we had something that other sites didn't have. We weren't
entirely sure what it was or what to do with it. Then maybe six months into the job it was
quite clear: look, we have to do something different with what we have here or it’s going to
go away.

Though all media, and therefore any given internet site, have always been “social” (Papacharissi,
2015), Friendster was positioned as an early example of what is recognized as a social media platform
today. Importantly, Friendster was also one of the first (commonly identified as the first) of this style
of site to emerge after the dot-com crash of 1999-2000, in which internet-based companies with
high economic valuations seemed to shutter overnight (Sapnar-Ankerson, 2018). As such, internet-
based company valuations that followed the dot-com era were haunted by the crash’s vagaries of
economic valuation.
My focus in this project is on companies like Friendster that came after this crash, but
interviews that I conducted with employees who worked on social websites prior to the crash, like
GeoCities, are elucidating in understanding the indeterminate nature of valuation as Friendster
began. GeoCities was founded in Los Angeles in 1994; it allowed for the building of personalized
web pages which were organized into interest-based neighborhoods: Hollywood was for discussing
films and actors, Heartland was for parenting and family topics, TimesSquare was for computing

79
and video games, and so on. GeoCities was one of the companies with massive valuations before the
crash, and they were purchased by then dominant internet giant Yahoo! in early 1999 for 3.5 billion
dollars in Yahoo! stock. Leading up to its purchase, the GeoCities organization knew where its
economic value came from, and knew what components could raise that valuation ever upward.
Warren, who worked at GeoCities through its acquisition to Yahoo!, explained it as follows:
Valuation was [based on] attention. The terminology that we used a lot in that time was
eyeballs, like: how many eyeballs did you get on the site? There was advertising, we had
banner ads, and stuff like that. But [advertising] wasn’t really a primary focus of what we
were trying to do, honestly. It was just growth. Yahoo’s valuation of us, I believe, was a lot
about just the size, the audience that we’d be able to bring, keeping it up. It was just really
important to be able to say we were the number one most visited website in the world.

GeoCities is representative of the modes of valuation endemic to the dot-com era. Eyeballs, or the
size of audience and their number of visits when compared to other sites, was what signaled value,
not the actual dollar amount that was currently brought in through advertising. While these modes
of economic valuation to some degree continued after the dot-com era–it continued to be better to
have more users rather than few, while more users continued to mean more potential for eventual
profit–the number of ‘eyeballs’ following the crash did not necessarily represent the value they did in
the dot-com era. What this suggests is that Friendster’s value indeterminacy was caused in part by
shifting classifications about what value was comprised of for “social” internet-based businesses.
An article in Fortune discussing Friendster’s rise signaled as much, describing how three years
after the dot-com bubble had burst, the internet had once again gained economic credibility
(Heinemann, 2004). Nonetheless, how one translated newly developed modes of valuation to the
social networking context was yet unclear: “Thanks to an assortment of thriving e-companies […]
Internet business models now exist. The question is how, and whether, Friendster can apply them to
its brand of social networking,” the article read. Indeed, the indeterminacy over proving out
Friendster’s value in the wake of the dot-com crash tracks with the internet economy’s

80
transformation in general, as technologists and venture capitalists in this era went from focusing on
value that could be realized through ecommerce and paid advertising to value that could be realized
through information accumulation about users through data tracking, data that could in turn be sold
to third party data brokers or employed as valuable market insights that would attract advertisers
(Myers West, 2017).
Friendster would eventually go on to be a cautionary tale for other platforms who needed to
articulate why they had value to both users and investors. As developer Jerome noted, today startups
need to have a simple and definite value proposition: “They just want to hear that you can distill
what your core value is very, very quickly. I think a lot of that comes from PTSD from Friendster
and a lot of other startups that were there at the time, that just never really clearly articulate what it
was they were delivering.”
Indeed the company and its investors didn’t know it then, but the first year of Friendster’s
existence would represent the high point for its economic valuation, even with indeterminate metrics
for what value meant. However, as Friendster started to enter what the organization largely
understood to be decline, the metrics for what constituted value and which components of the
platform possessed value became much clearer in the eyes of the organization.
When Does a User Group Possess Value?
Friendster appeared to be a Silicon Valley company through and through: created by a
Silicon Valley entrepreneur, first serving to connect Silicon Valley’s young tech workers, and led by a
board that included well known Silicon Valley venture capitalists. Yet, as employees explained, a
significant part of Friendster’s userbase was based not in Silicon Valley, nor even in the United
States. Rather, the platform became popular with Southeast Asian users, especially in the Philippines,
in part because a famous Filipino actress in San Francisco started using it, as architect Peter
recounted, but also because it fit in with the tightknit friend groups of Filipino classmates known in

81
Tagalog as “barkada,” international development manager Steve explained. Employee estimates of
the ratio of Southeast Asian users to US users ranged from 3:1 to 5:1.
As employees noted, this user community did not align with the long-term vision for the
platform or the organization’s classification as to which users had value. It was a sentiment
particularly held by the board and other company leadership. Instead, the users in Southeast Asia
were thoughts of largely as possessing negative value: they were a “problem,” a “nuisance,” or an
“annoyance,” and ultimately by some, understood to be Friendster’s catalyst for decline. From the
beginning, Friendster employees were under the directive to focus on US users and attempt to
reduce the impact of the Southeast Asian users, lest these communities curb the platform’s growth
in the United States, and therefore, its potential for monetization and its trajectory as a runaway
success. Steve, the international development manager, summarized the board’s directive to the
Friendster staff as follows: “Your mission is to help Friendster in the United States. That’s where the
revenue is. That’s where our biggest competitive threat is. We are not a Southeast Asian social
network.” For those who understood Friendster as a social network for US users, the slowing of
growth in the US and the dominance of the Southeast Asian users signaled Friendster’s decline in
value, even if there was growth in the total number of users. Kelly, a product manager, described the
moment where it was realized the Southeast Asian users were Friendster’s primary community:
When you start to dig into that problem set, then you get to this thing, which is like,
‘Oh, a tremendous amount of usage is overseas.’ We’re taxed with all the burden of
serving this audience, which is massive and super resource intensive, and there’s no
monetization potential here.

Continuing, she compared Friendster’s decline to the demise of Google’s Orkut platform,
which had lasted from 2004 to 2014:
[Orkut] had been essentially decimated by… It became the platform of choice for a whole
bunch of Brazilian soccer hooligans. […] This is another lasting lesson for me coming out of
this as a product person, as a marketer, as a community person, is like, the community

82
chooses you. That community chose that platform, and it strangled the platform. They
couldn’t attain a US audience because anytime someone would log in everything going on
was in Portuguese and it was about soccer.

One of the ways in which users in Southeast Asia were framed as having negative value is the
characterization, as seen in Kelly’s account, that they were “resource-intensive,” requiring significant
computing power to handle the number of users coming online with what Friendster’s board
understood as comparatively little potential for monetization. Friendster had indeed invested
significant funding into sophisticated technical capacity to handle all its users, as one of their early
and significant challenges was keeping the website up as traffic poured in. Employees from the
2004-2005 era discussed Friendster’s purchase of a Hitachi SAN (Storage Area Network), what
CEO Scott Sassa described as the “the most expensive hard drive in the world,” to adequately
handle their large userbase. They also lamented that US users were dissuaded by consistent crashing
of the site, while their international users did not seem to mind and returned to the site regardless.
(The perceived differential between US-users and non-US users tolerance for website instability may
have been a misrepresentation, as other Friendster employees noted that MySpace also went down
frequently without affecting US users’ desire to return to the site.)
It is not a given that users with less monetizable potential would automatically indicate
negative value for a company. Kelly described how, following Friendster, she had worked at AOL
on their chat products, like AOL Instant Messenger (AIM). She noted how the value that the
platform provided to its users was front of mind when changes to the service were proposed:
Whenever we wanted to make changes to that, they were like, ‘But there's this chat room of
Scottish lace knitting grandmas who have gotten together every Saturday night at 8:00 PM
for 15 years, and if we shut this down, we’ll break all those relationships.’ It was extremely
user focused. So, in my experience, it really can go either way. It’s where the company thinks
the value is, and what about their users they value.

83
Understanding that users can possess value outside of their potential for economic extraction, the
sense of Friendster’s Southeast Asian users as being a burden to the organization–possessing
negative value and being resource-intensive–reveals how users have differential value in the eyes of a
holding company. For Friendster, not all clicks, eyeballs, communities, and signups held the same
worth. Moreover, this is a differential value that appears to be deeply xenophobic, with company
leadership not only indifferent but unwilling to see how non-US users could possess any value at all.
International development manager Steve, for instance, noted that he had to hold meetings with
developers and product managers after the CEO would go home. He called these meetings
“Features after Five.” There, he would work with employees to build out features and brainstorm
monetization opportunities for their non-US audience.
As stories like Steve’s indicate, not all Friendster employees felt that these users indicated
negative value. Some looked to these users as a sign of vitality instead of as a burden, in terms of
both monetization and from a community perspective. This sentiment just wasn’t necessarily
reflected at the board level. Tom, a developer, explained leadership’s eventual change in thinking:
At first it was like the Philippine users were considered to be an annoyance. Like, ‘These
people are here, they’re crushing us, we gotta get the US back.’ Then at some point they did
pivot and were like, ‘Wait–these Philippine users are our users. There’s a lot of them. Let’s
figure out how we can serve them better.’ But I think it took a while for them to make the
transition from having fallen from the top of the US to being like, ‘Okay, well maybe there
are other things we can do.’

Sivu, who served as CTO from 2006 through its sale to MOL, noted that investors Kleiner Perkins
had indeed nearly shuttered Friendster after its primacy in the US market had faded. They decided to
try again, however, with the knowledge that both Southeast Asia user numbers and those markets’
economic potential could grow. As Sivu noted:
But then [Kleiner Perkins] said–those days, they had eight million unique users in
Southeast Asia. Would you shut down something which had that many, eight million
users? […] We believed that they would get more than eight million users, if they resolve

84
the technical problems. That is why they didn’t shut it off.

Sivu and Tom’s description of how Friendster leadership eventually recognized that even Southeast
Asian users could spell economic value illustrates a shift in valuation as company leadership
understood the platform as in decline. This transformation in the valuation of user groups went
hand in hand with other changes in value, as where value resided and from where it could be
extracted became more concrete in the eyes of the organization. I discuss these changes in the next
section.
Extracting Value
Having acknowledged that US users were largely a lost cause for revenue, and that there was
potential for profit by focusing on the platform’s sizeable audience in Southeast Asia, the company
went looking for ways to extract value from platform components that had otherwise been
backgrounded during their ‘successful’ early years. As they did, the Friendster organization appeared
to engage in practices that resembled the act of salvage, shifting their valuation strategy from
understanding Friendster as a wholistic platform system to a platform that was composed of
subdivided components that individually possessed potential extractable value.
The primary means of doing this was by selling their user audience to someone who wanted
to focus on a Southeast Asian market. Richard Kimber, a Friendster CEO based in Australia and
who joined the company in 2008, described the directive as such: “What the chairman [of
Friendster’s board] said to me was, ‘Fix it and flog it.’ Simple.” Fidelia, a product manager who had
worked at the company just prior to Kimber’s hiring, had described a similar injunction. Employees
needed to fix up Friendster–especially its revenue and user numbers–such that they could attract an
interested buyer:
I don't know that we were looking at the super long-term strategy. It was more–and this is a
terrible way to put this–like putting lipstick on the pig. I think the goal was to find an

85
investor, or to get acquired, and to show people that Friendster is still a valuable brand, that
[there’s] revenue potential here.

In the years between 2008 and their eventual sale to MOL in 2011, Friendster had conversations
about potential acquisition with a variety of organizations: Globe, a Philippine telecom company;
another company in South Africa; and their eventual purchaser MOL. Sivu recounted how
Facebook had also expressed interest in purchasing them, acknowledging that they would “never be
able to get that kind of traffic in Southeast Asia. We were too big, so they would just buy us for the
traffic.” Employees I spoke to from this era had conflicting opinions on what path Friendster
should have taken from this juncture, and if MOL was ultimately the best purchaser: nonetheless,
they acknowledged that the primary asset they were selling was their large and engaged user audience
in Southeast Asia.
The second major component Friendster salvaged was their intellectual property in the form
of patents. One employee described how she had always heard “a vague whisper in the corner about
something about patents,” but she–nor anyone at the company–had ever made much of it.
Friendster had begun quietly patenting broad aspects of social networking in 2006 under its then
CEO Kent Lindstrom, as the platform began to go into decline (Gannes, 2010). They included
patents for actions as broad as “making connections on a social network, friend-of-a-friend
connections through a social graph, and social media sharing,” as an article describing them
enumerated (ibid). As Friendster sought to extract value from its un-bundled components, however,
these patents became what an employee later called the company’s “crown jewels.” These patents
had the potential to threaten other social networks, especially Facebook, should those companies go
public. Indeed, as Friendster was searching for its buyer, Facebook was preparing for its eventual
IPO, which transpired in 2012. To have a seamless public offering, Facebook needed to purchase
these patents, which they did in 2010 for $40 million dollars.

86
When Friendster was purchased by MOL in 2011, the site’s Silicon Valley investors had
extracted what monetary value they could by divvying up the platform’s components and selling
them to the highest bidder. Following the acquisition, Friendster’s existing employees in the US left
the organization, and MOL transferred physical assets to their headquarters in the Philippines. In
fitting with the sea-faring roots of salvage, these servers made their way there by ship.
However, what is left out of this discussion of components that possessed value, that had
the potential for salvage, are the pieces of platform that were deemed to be rubbish. These may
indeed be the components that were not brought up in interviews, components that were hidden
away or made invisible by virtue of their value dispossession (Thompson, 1979). While user content
was included on those servers, this content seemed to possess little obvious value to the holding
company; it was deleted by MOL in 2011 as they rapidly transitioned away from a social network
and towards a platform purely for gaming. (Former CEO Kimber noted that MOL had likely done
this to “take away some of the complexity” and “cut the cost” of running their new acquisition; the
company, who was not in an advantageous financial position, had been pouring significant money
into the platform, and he conjectured that they had not been prepared for the engineering
complexity that social networking presented.) Employees also rarely discussed how these troves of
user data could be mined for insights about users, which became a major component of value
extraction on later networks (Myers West, 2019), including the next case I examine, MySpace.
12

MySpace: The Rise, Fall, and Attempted Resuscitation of Value
It is fitting that MySpace should follow Friendster in this chapter, as so many of my
interviewees from both platforms discussed how MySpace followed on Friendster’s coattails.

12
In our interview, MySpace employee David mused about how a focus on user data seems to determine
which companies survive and which don’t. “I'll say this, I don't think the people who created Friendster saw it
in terms of user data. And I don't think the people created MySpace thought of it as user data. And that's
probably why they're dead. Right? [That’s] difference between [them and] Facebook…which changes
everything, right?”

87
MySpace was the social platform popular in the US in the mid-aughts: the most visited social
network from 2005 to 2009 (Moreau, 2022), the most visited website in the US in 2006 (Reuters,
2006), and during its heyday the largest repository for music online.
The social network launched as an ad-hoc project of Los Angeles-based company eUniverse
in 2003. eUniverse itself began in 1999. Prior to MySpace, their business was email marketing:
accruing, spamming, selling, and selling to millions of email addresses. Addresses were collected
through a network of websites, whose foci ranged from animated e-cards to inspirational quotes
with a “native American theme” (Hansell, 2001)–thus, like Friendster, encoding normative
whiteness, in this case through the “long-standing tradition of white hegemonic control over Native
identity” through the commodifying of that identity, as Jason Edward Black has put it (2002, p. 606).
As a New York Times article would describe (Hansell, 2001), eUniverse’s email accrual sites “are
cheap, corny, crass and profitable.” In other words, eUniverse’s business was what might typically be
called “trash” of the internet, in their production of “junk” mail, the go-to metaphor for spam email
(Brunton, 2013); or in the production of what is often considered low value, “trashy,” click-bait
content, as the Times article evinces (Hansell, 2001). Moreover, the company was built on market
salvage. As Aaron, an early board member at eUniverse explained to me, company founder Brad
Greenspan had launched eUniverse by purchasing a defunct but publicly traded company called
Motorcycle Centers of America. Using the shell of Motorcycle Centers of America, eUniverse–
which renamed itself to Intermix Media in 2004–built its offerings across its numerous sites and web
services.
In its early years, eUniverse was interested in expanding its business by coming up with new
products that the company could sell to their email lists; this included things like printer cartridge
refills, anthrax test kits, diet pills and CDs. They also thought to expand their business by accruing
more email addresses, which included the purchase of another email marketing company just down

88
the road from them in Los Angeles. Called ResponseBase, it was run by Tom Anderson and Chris
DeWolfe–MySpace’s eventual founders. With millions of email addresses of their own,
ResponseBase was easily absorbed into eUniverse. Though Anderson and DeWolfe had stayed on to
help the transition, it was only a short time later when they were called into CEO Brad Greenspan’s
office. Greenspan wanted them to come up with another opportunity under the eUniverse umbrella,
another way of generating more email marketing revenue.
13
Moreover, as board member Aaron
recounted, Anderson and DeWolfe also needed to come up with a strategy to generate more
revenue to satisfy the terms of their acquisition agreement with eUniverse.
Their idea was a social network: in essence, another node through which to collect email
addresses. Recognizing the success of recently launched Friendster, they thought they would put
together a “rock ‘n’ roll social network,” as an early employee called it, one that enabled bands to list
their gigs and invite their fans, thereby accruing more contact information. They decided to build it
using a domain that eUniverse had recently purchased–MySpace.com–that the company had been
using to sell $99 e-scooters (Hansell, 2006).
14
As this eUniverse employee told it, they put together
MySpace rapidly, launching it internally for eUniverse, then regularly sending invites to their millions
of email addresses, and finally growing without these prompts as the site became well known
beyond their email lists. MySpace soon became eUniverse’s main focus.

13
There was a ripple of controversy around MySpace when a 2006 article on the now defunct, Gawker-
owned, Silicon Valley blog ValleyWag (Lapinski, 2006) reported that MySpace’s founders were originally in the
business of spam, adware and pop-ups, and that MySpace had grown in part because of its aggressive email
marketing. As a New York Times article reporting on the controversy noted, to industry insiders this was not a
new revelation that MySpace’s founding team was “not a band of do-gooders” (Baker, 2006; quoted in
Mitchell, 2006).

14
Prior to the purchase of the MySpace domain name by eUniverse, ‘myspace.com’ was owned by MySpace,
a cloud storage company. (An article chronicling their demise called this service an “online closet” [Morris,
2001]). In a sign of things to come, when MySpace the storage company went under in 2001, it deleted
“untold numbers of files stashed there free by its registered users,” having given its users six days’ notice to
retrieve their files (ibid). The defunct cloud storage company estimated that they had deleted 7.5 million files
as they sunset.

89
As Friendster developer Jerome described earlier, Silicon Valley companies typically benefit
when they are able to “distill [their] core value […] very, very quickly.” He gave an example of what
it looked like to do so: investors, he noted, want to hear things like “I’m the Instagram of rock-
climbing,” or the “Google of LinkedIn.” Distilling value meant having a simple pitch that expanded
on the proof of concept for a business that had already gained market traction and public
recognition. This is indeed how MySpace explained itself to potential employees and how they
explained their concept in their pitch to eUniverse leadership. Louis, who managed database
operations for MySpace for several years, recalled interviewing for his position in 2004. Not yet
familiar with MySpace, the founders conceptualized the site for him as like Friendster, but with
music: “They looked at Friendster and said, ‘Okay, that's cool. Bands should be able to do
something like this.”
This “value-add” mindset that eUniverse brought to the development of MySpace, in which
they iterated on Friendster as a social networking foundation, is ultimately framed around
concurrent value of the utility the platform would provide to its customers and the economic value
it could realize for the company by exploring an otherwise untapped market niche. As I’ll explain in
the next section, MySpace employees and the MySpace organization articulated two kinds of value
that they believed the platform possessed as it grew: economic value and cultural value.
Peak Economic and Cultural Value
As MySpace grew from an ad-hoc project to a dominant social media platform and one of
the most visited sites on the internet, it accrued significant economic and cultural value. Its rise in
economic value is defined by two events: its acquisition by Rupert Murdoch’s News Corp for $580
million dollars in 2005, and a $900 million dollar deal with Google in 2006 (Van Duyn & Waters,
2006). News Corp purchased Intermix, MySpace’s holding company, then rapidly shut down its
ecommerce verticals: it was only interested in MySpace, which represented a bid for next-generation

90
media audiences. While the over half a billion-dollar acquisition was understood as an aggressive and
sizable purchase, it did not necessarily appear that way for News Corp. Aaron, the board member,
explained how that organization felt they got MySpace “for a song,” having immediately made a bid
for digital relevance through their acquisition of the number two most visited site in the world.
MySpace at that time had “no end in sight.” In their acquisition, News Corp chose to keep on
MySpace’s two main founders, DeWolfe and Anderson, to steer the company and maintain the site’s
connection with its youthful userbase that included the 18–34-year-old demographic, a highly
desirable one for advertisers. As an employee familiar with MySpace’s business plan noted,
“monetizing that audience is very valuable.”
In 2006, MySpace also struck a deal with Google, who had beat out rival Microsoft to supply
MySpace’s search function and keyword advertising services. At the time that this $900 million-
dollar, four-year deal was negotiated, the site was growing by a quarter of a million users per day
(Van Duyn & Waters, 2006). Google had struck this exclusive deal certainly because it was a major
play for users’ recognition in a crowded search engine market, and certainly because it would mean
revenue through the ability to serve ads on their site. Importantly, however, the Google deal also
allowed them to gain insights about users through data accrued through the use of search and
advertising, thereby allowing better ad targeting for the company in general. Thus what gave
MySpace its economic value was not only massive traffic and the potential for advertising “eyeballs,”
moreover the “right” kind of eyeballs, but also because traffic had started to signify valuable data.
MySpace thus acted as a microcosm of the internet economy’s transition more generally, a time
when speculative and realized economic value went from being signified by numbers of ‘eyeballs’
that might see an ad to data generated by users visiting the site (Myers West, 2017; Sadowski, 2019).
MySpace’s roughly year-long period of acquisition, rapid growth, and advertising revenue deals was
described to me as the “honeymoon phase” by one employee. The expansion of its user base along

91
with its realized potential for advertising revenue seemed to illustrate the sizable economic value of
social media, a type of value that Friendster had projected but had been unable to fulfil.
Aaron summarized how value was constituted in the acquisition: “News Corp, they saw it
strictly from a money-making perspective. How do we turn this into a revenue machine?” For the
employees I spoke to, however, MySpace represented more than its economic potential. Instead,
employees discussed how MySpace possessed cultural value. (Employees’ emphasis on cultural
instead of economic value may also be because MySpace did not portend significant money for
employees outside of their salary; because of the configuration of stock options, most employees did
not strike it rich when MySpace was purchased by News Corp). When I say “cultural value,” I draw
on definitions from cultural studies and sociology to mean the relative worth of an entity within a
particular social milieu, and that entity’s (or affiliation with that entity’s) ability to effect change
(Oakley & O’Brien, 2015). This sense of cultural value–at least as it was understood by employees–is
related to what sociologist Pierre Bourdieu has called “symbolic capital” (2013), the distinction of
prestige, legitimacy, and authority that a person possesses, and which operates within a larger social
system of cultural boundary marking. The cultural value, and therefore the cultural legitimacy, of
MySpace was highlighted through stories where interviewees recounted how being affiliated with
MySpace in its peak years made them feel like a “rock star,” a person defined by their cultural
renown, especially within the realms of entertainment that MySpace personnel understood
themselves to be a part of. Louis described 2006 as follows:
Everywhere you go, people would be like, "Oh my god, you work for MySpace?" […] It was
the whole Hollywood lifestyle. You could get into bars with your [business] card, you get
free drinks, people are telling you how great you are, how wonderful it is. Total ego stroking
for an entire year. It was amazing.

Brendan, in turn, had noted how when he would wear his MySpace sweatshirt to the grocery store,
people would treat him like a celebrity, asking: “Do you work at MySpace? Oh, my gosh. Can I take

92
a picture with you?” Similarly, Aaron noted that at MySpace the “brand culture” was also the
“company culture,” meaning that the lifestyle that MySpace proffered as a brand–as a nexus for the
world of music, entertainment, and celebrity–was the lifestyle that employees and company
leadership embraced. (danah boyd [2011, p. 206] characterized MySpace’s brand culture as one
“tightly entwined with its symbolic reference to maturity, status, and freedom in the manner
espoused by urban late-night culture.) Numerous interviewees discussed how being at MySpace
during its peak years was thrilling, with the regular spotting of celebrities in the office halls and
elevators. This legitimating power of working for MySpace, its ability to hold symbolic capital for
employees, was one of the major avenues through which they understood the platform’s worth. In
turn, when MySpace began declining, that decline became evident for employees also through signs
of the site’s waning cultural relevance.
“The tools we had to fight were not working anymore:” Decline in Cultural and Economic Value
By 2008, the honeymoon phase had ended. From the perspective of the News Corp
organization and the perspectives of MySpace employees, a tension emerged between economic
valorization and cultural valorization. In interviews, employees described how News Corp wanted to
ensure they made good on their investment in the site, aggressively monetizing the site such that ads
cluttered pages and brand collaborations were done with little regard for that brand’s cultural cache.
As Leah, a MySpace employee who was crucial in getting hip-hop artists to join the site in its early
years, noted that after the News Corp takeover, the organization’s economic interests conflicted
with the cultural interests that employees sought for the site:
We became slaves to the advertising industry. In the advertising industry, it’s not the coolest
brands, it’s the ones with the biggest money: McDonalds, Coca-Cola. [What] are the
products that we're pushing to these younger folks? It was those brands. That’s how my job
changed. It changed from really cool and editorial to then, “How do we make the most
[money]?”

93
As Aaron explained it, News Corp wanted to get to a billion dollars in realized revenue per year just
for MySpace, and as such had their sales team–traditionally focused on advertising for their other
print-based and TV-based verticals–focus on talking to advertisers for this digital property.
“MySpace now became one of the things that they sold,” Aaron noted. In turn, the user experience
changed: “As a user, you were stuck seeing whatever the guys at News Corp had sold. I think the
user base started to rebel at that point.”
What’s more, the lucrative deal that MySpace had made with Google in 2006 had few
prospects of being renewed under the same beneficial terms. When a deal was renewed with Google
in 2010, it was indeed a much less profitable agreement. An employee familiar with the deal noted
how when it was renewed, Google had already had years of experience monetizing MySpace’s ad
inventory, and thereby had accrued significant amounts of valuable information about user behavior.
As such, Google was not incentivized to pay a premium for serving ads there, as the data that could
potentially be accrued was significantly more redundant.
At the same time, Facebook had gone from what employees at first considered just a
“college social network” to one that was rapidly moving to accept all users. Facebook was defined
by a consistent interface–what employees often called a “clean” interface–along with the use of real
names as opposed to screen names. Employees contrasted Facebook’s pared down design with
MySpace pages that were ‘cluttered’ with ads and could be customized at-will by users with
backgrounds, songs, and widgets. Moreover, the growth of Facebook and apparent waning of
MySpace was influenced by racialized and class-level dynamics, what boyd (2011) analogized as
“white flight” from MySpace to Facebook as young, affluent white people tended to move to
Facebook while poorer youth of color tended to use, and were associated with, MySpace––at least in
this era of transition between the sites. While boyd’s analysis (2011) of the racialized movement of
users between Facebook and MySpace does not address how these dynamics influenced the site’s

94
economic value, and MySpace employees in turn did not address racialized dynamics of user
movement in interviews, it would not be surprising if the perception of MySpace’s decline was
influenced by organizational understanding that these users with lower socioeconomic status were
‘lower value’ users when compared to their affluent and white counterparts.
The means by which employees came to understand MySpace’s decline in the 2007-2009 era
were indeed complex, but the most tangible evidence of decline for the organization was through
user metrics. Digital metrics (alternatively, digital analytics) include information like page views,
amount of viewing time, number of comments, posts, or likes, and paths through numerous pages,
measurements describing user behavior as users interact with a site (Tandoc, 2014). These metrics
are in turn used to locate value and attach meaning to user behaviors on a site (Beer, 2017). The
importance of digital metrics in media production cultures has been discussed most in the context of
online journalism, wherein figures around audience engagement have been shown to shape the
editorial process at numerous stages (Beer, 2017; Christin, 2020). Research on metrics and social
media platforms shows how analytics displayed to users–like the number of likes on Instagram or
the number of retweets on Twitter–act as markers of social distinction (Paßmann & Schubert, 2020),
while it is also known that user metrics influence the design of algorithms on platforms, which in
turn shape user experience on those sites (Couldry & Powell, 2014). Despite these studies, there is
limited empirical evidence showing how social media platform employees have made sense of user
metrics within organizations. While this study does not focus on metrics singlehandedly, as I’ll
discuss here it does point to the significance of digital metrics for internal comprehension of a
platform’s value and its competitiveness with like entities, especially when this value appears to be
shifting.
The importance of the hockey stick to visions of platform success is a case in point. For the
employees I spoke to, the hockey stick most often represented a basic metric: new user sign ups.

95
Decline was glimpsed through similar basic metrics, typically the waning of new user sign ups, as
well as the total number of users and total minutes spent on a site when compared with competitors.
In 2007, 2008, and 2009, MySpace consistently compared their metrics with public metrics about
Facebook. The rise of Facebook user numbers and the slowing of MySpace user sign ups and user
engagement signaled to MySpace employees that they were in decline. “I recall watching the number
of signups, and how many new signups had come in, and tracking those numbers day in, day out,”
developer Doug recounted. Albert noted how in that era employees paid attention to Alexa, the
ranking of top websites, and how MySpace was faring against Facebook. “We paid a lot of attention
to that back then. As soon as we saw those numbers shift on there, it was, ‘Oh, shit. This isn't
good.’” It became clear to employees through these metrics that Facebook was, as one employee
recounted, “on that growth spiral up. They were growing so fast. I was on Facebook. I was on as a
user. I could see the difference. It just seemed like the trends overtook us. Disruption happens in so
many industries. We got disrupted, and the tools we had to fight were not working anymore.”
Along with decline in user numbers that signaled the waning economic value of MySpace,
other qualitative measures illustrated the site’s decline. These were ones employees often phrased as
‘clear indicators’ that the company was not doing well: changing CEOs, as MySpace founders
Anderson and De Wolfe stepped down from their operating roles in 2009 and were replaced with a
new executive team; massive layoffs–going from 1,600 to 1,000 employees; and the closing of
satellite offices (Gillette, 2011). MySpace attempted an ill-fated redesign in 2009 as well, attempting
to provide more of a Facebook-like experience.
For employees who had been there through MySpace’s peak, there was also an obvious
difference in its cultural value. Brendan, who had earlier described how wearing his MySpace
sweatshirt to the grocery store would elicit attention from fellow shoppers, noted that after its
decline “I didn't wear any of my MySpace gear anymore. To be honest with you, I downplayed it. I

96
wasn't really telling people I worked at MySpace. That's how bad it got.” Similarly, Albert noted that
being associated with MySpace went from a badge of honor to a red flag. He described how:
It went from people having the utmost respect for anybody who worked at MySpace, to
people looking at you side eye: ‘Are you really good? Because you used to work for
MySpace:’ Not appreciating the engineering side of it, the tech that was behind it, but just
focusing on a tarnished and dying brand.

Through 2009, as Louis recounted, employees still believed the site could be turned around. “We
were still salvageable,” he argued, noting that the site “still had huge amounts of value” for celebrity
and music promotion. In 2010 and 2011, however, MySpace employees who had held on since the
beginning acknowledged that, as Louis explained, “it was over.”
Salvaging Value Through Data
As MySpace struggled to find ad deals in 2011, especially ones with long term contracts
(Steel, 2011), News Corp also agreed that MySpace was over. They went searching for a buyer for its
now diminished asset, and eventually found one in Specific Media, a digital advertising company
based in Irvine, California. Specific Media specialized in placing ads on remnant online ad space,
meaning that it brokered cheap online ad space when websites could not fill these slots with higher-
paying advertisers. Specific purchased MySpace for $35 million dollars, with the idea of relaunching
the network as a music artist portal. The vision for it, as it was told to me, was somewhere between a
Spotify, the popular music streaming app, and a social network. In 2013, MySpace completely
relaunched, transferring existing user accounts to the new platform, a process I describe in the next
chapter.
Both News Corp and Specific Media strategized around extracting final monetary value out
of MySpace. News Corp no longer wanted the burden of a diminished, non-revenue generating
online property, and extracted the last $35 million they could from it, while Specific Media saw the
opportunity to salvage specific parts of MySpace and refashion a revenue generating entity from

97
these components. Julian, a content writer who worked for MySpace during this transition,
described the value-related reasoning of the Specific Media organization purchase of MySpace:
They bought this thing at a fire sale and I think they realized that like, ‘We bought a brand
name, we bought a lot of user data, maybe some emails that may or may not be active
anymore, and then we bought a thing that once had this really vital cultural relevance to a lot
of people.’ I think the guys from Specific Media really latched on to the fact that they had
this brand name and then they had all this data, [and] were like, ‘All right, we could do stuff
with that.’

Julian summarized here what leftover value was identified in MySpace: its data, its brand, and its
one-time significant cultural cachet. These were components that Specific attempted to reconfigure
into a whole that could revivify its economic and cultural value. The employees I spoke to reasoned
that the primary motive for Specific’s purchase of MySpace was for data. In contrast to Friendster,
for whom user traffic was framed as a major asset in its purchase by MOL, as an advertising
company in an evolved advertising market Specific saw user data as a significant asset. They saw
value in both the accumulated data from the historical MySpace as well as the potential generation of
new data from the reconfigured platform. David, who steered creative content production under the
Specific transition, explained the two types of data that were envisioned as valuable:
The thing [Specific] always touted was ‘We have user data from x million amount of users.’
A lot of that data is outdated, so I never paid it any attention. What I didn't realize until years
later is that even if that data is outdated, it’s still useful to ad sales companies because it helps
build profiles and model behavior.

Indeed, while Google’s ad network had had access to years of MySpace data through their earlier
deal, Specific was purchasing data that was new to them and their ad network.
15
At the same time,

15
This recalls Oscar Gandy’s (1996) description of the “panoptic sort,” or the ways that public
surveillance is enmeshed in corporate structures. As he writes, in these contexts “all information
about individual status and behavior” is seen as “potentially useful in the production of intelligence
about a person’s economic value” (p. 133), just as the ‘historical’ information possessed potential
value for Specific’s ad network. Importantly, Gandy discussed these logics in the context of direct
mail and other offline media, showing how digital platforms are not the first or by any means the
only entities to generate these logics.

98
David also noted that the rebuilt platform could serve as a node through which to generate new data
assets:
The other thing I didn't know then that I learned after I left [was]: they just needed users.
Because users provided data. And that data would become very valuable. One of the plays
they wanted to make, because they were a music specific platform, was they wanted to sell
that data back to the record companies. I don't know if it was an immediate business need,
but they felt like if they had insight into people's listening habits and user connections, that
became data points that can then be monetized.

Michael, another person working on the content team in this era, noted that these data assets would
be generated by luring previous users back into the ecosystem by providing them with free music.
As such, they could also generate data that would allow them to refine their ad placements and serve
that arm of their business. The logic, he noted, was that “once they interact with MySpace, we can
stick this little tracker on them and follow them all over the internet and then that’ll help our ad
business.” One of the ways that MySpace thought they could lure people back in 2011, according to
interviewees, was a tactic that mimicked the one that MySpace first used to grow their site in 2003.
In both 2003 and 2011, MySpace leadership reasoned that they already had an existing connection
between the corporation and a large audience. In 2003, it was through accumulated email addresses.
In 2011, it was through existing accounts. Hundreds of millions of user accounts continued to
persist on the site, even if they were not frequently updated. They were viewed as an asset because
they provided a ready-made linkage between the platform company and a user audience.
Crucially, MySpace leadership separated the connection to users from the cultural production
of these users when considering these existing accounts as assets. Categorizing accounts as valuable
through their user connections instead of their users’ content was a means of separating the wheat
from the chaff, or perhaps more accurately, salvage from junk. Identifying value as residing in
subdivided socio-technical components of what was previously thought of as an integrated system,
under Specific Media company leadership attempted to rearrange these components into a new

99
whole. Like in Friendster’s case, user-generated content was largely an afterthought when engaging
in acts of platform salvage.
Conclusion: How Valuation Structures Commercial Digital Loss
This chapter has examined what happens to social media platforms after the hockey stick,
what occurs when one continues to track this diagram further through time. It looked to the ways
that value is identified and constructed by organizational leadership and employees, and the ways
that shifts in value–especially economic and cultural–come to connote a platform’s decline within
culturally-specific systems of valuation. These culturally-specific systems of valuation are aligned
with the neoliberal socio-economic structures of Silicon Valley, in which the worth of an entity is
determined within a context of unencumbered markets, free trade, competition and individual liberty
(Harvey, 2007). These socio-economic structures are most obviously aligned with economic
valuation during decline, but they also subtly shape the cultural valuations that occurred as well. This
can be seen especially through MySpace, as employees anxiously conjectured about the impact that
devaluation would have on their individual employment prospects. Employees feared being left
behind as the social media platform industry moved on.
In discussing the diminishing, salvaging and reconfiguring of value, this chapter describes the
landscape of value through which user-generated content moves during the process of sunsetting.
As such, it provides context for digital loss that often occurs during this process. This deletion or
removal of content is necessarily concurrent with its diminished value in a culturally-specific system
of valuation. Because the landscape I’ve described here is most concerned with economic valuation
within a neoliberal marketplace, the type of deletion and content removal that occurs can more
accurately be called commercial digital loss. Commercial digital loss can be conceptualized as the
removal of user-generated content by the corporate channels hosting this content. Indeed, both
Friendster and MySpace’s cases show the commercial logics that influence understanding at the

100
production-level of when an organization is in decline, what caused this decline (see also Corry, 2021),
and which components can be salvaged for value or left for rubbish.
Along with the commercial logics that structure valuation during sunsetting, these cases also
reveal how existing forms of domination along socioeconomic, racialized and geographic lines shape
which users are understood to possess value, and even which users are considered as having no or
negative value, both economic and cultural. Anthropological writing about waste has shown the
ways that junk and rubbish are slippery terms that become transferred through metaphor to other
domains, especially the description of what a dominant group may label as ‘superfluous’ people who
they deem to provide no or negative societal value (Eriksen & Schober, 2017). It is a metaphor has
been often mapped to refugees, to the unhoused or, in this case, to user groups.
Because there was little economic or cultural value identified in user-generated content
during these moments of decline, it could be asked: did user-generated content ever have value in the
eyes of organizations and in the eyes of employees at platforms like Friendster and MySpace? As
noted earlier in this chapter, it did when the production and storage of content was aligned with
traffic; when the storage of this content brought people to the platform and presented fresh
opportunities for advertising and data insights. These were moments when structures of valuation–
the social and cultural meaning of content to users, and the economic meaning of this content to the
organization–aligned. It is when these structures of valuation are no longer aligned that dominant
forms of valuation become clear, as they do in the moments of decline explored here, when the
dominant positioning of user-generated content as trash–or as having been emptied of realized or
potential economic value–allows this content to be trashed.
These value structures–the cultural meaning of user-generated content and the economic
meaning of user-generated content–are also commonly represented as aligned by platform
organizations when, in public facing statements, they laud the creativity, freedom of expression,

101
connection and self-representation that their system fosters by serving as an intermediary for user-
generated content (Gillespie, 2010). It is expressed through mission statements like Facebook’s,
where they “give people the power to build community and bring the world closer together,” or
YouTube’s slogan, to “Broadcast Yourself,” or Instagram’s, “Capture and Share the World’s
Moments.” What these slogans also describe are the ways that processes of valuation–determining
the worth of something–can be intimately related to values, or what a person or group of people
deem important in life (Friedman, Kahn & Borning, 2006). The interplay between valuation
described in this chapter and values can be described through something Kelly, the Friendster
employee, said in our interview: “It comes off sounding very cynical, but I think it’s the truth:
especially at this late stage of that company's life, people are just picking the bones clean. Once
they’ve taken everything that they think has value, nobody’s like…” She wailed jokingly: “But what
about the users!” While Kelly’s statement shows how dominant systems of valuation described in
this chapter were largely ones proffered by organizational leadership which shape the realm of
possibilities for sunsetting, in the next chapter I’ll show how values–especially of individual
employees–also influence this process.

102
Chapter 3
Social Values and the Shaping of Platform Afterlives
We were 45 minutes into our conversation when I asked Julian, a former MySpace employee, how
the organization had come to decide what to do with user data as they transitioned to the ‘new’
MySpace. Otherwise eloquent, he suddenly seemed to trip over his words:
It’s tough to say. I don't know. I don't know if, yeah. I don't know if… you know, who
knows. I don't know. I couldn't tell you. I think most of the people, I will say the majority of
the people there were like, we have to act, we have a responsibility to this stuff. Even though
we know we have to move on from it.

Discussing the particulars of sunsetting with employees tended to strike a nerve. These questions in
general brought reactions that were anomalous in one way or another: they resulted in answers that
conflicted with actions, responses tinged with negative sentiment toward the holding company,
descriptions that obfuscated the sunsetting process, and employees emphasizing their desire for
anonymity.
In this chapter, I frame these conversations as what Shilton and Greene (2019) have called
“values levers” in technology design (see also Shilton, 2013), where particular work practices in the
design process trigger discussions of particular values. For instance, in their study of mobile
application developers’ attitudes toward privacy, one ‘lever’ for addressing this value was engaging in
discussions about analytics. Here, I adopt the notion of a value lever analytically and deductively,
understanding levers as work practices (or the discussion of these work practices) that make
discursively visible the varied values employees bring to that work practice. In particular, I focus on
instances where employees grappled with how to handle user-generated content as platforms were in
decline or ultimately shut down. Throughout this chapter I use these moments, these often-
anomalous conversations, these levers, to explore the attitudes that employees brought to the
sunsetting process, and how these values come to shape digital memory infrastructure’s afterlives.

103
I look to two different cases here to parse values in the sunsetting process: MySpace, the
mid-aughts profile-based platform; and Vine, the short form mobile video app owned by Twitter. I
focus on these sites in both this and the following chapter for reasons of access and comparison. In
the cases of MySpace and Vine, I was able to speak to employees who had specific knowledge of the
organizations’ treatment of user content over time; other major sites that I’ve discussed so far, like
Friendster and GeoCities, had fewer employees who could speak directly to those practices. At the
same time, these two platforms provide fodder for comparing values in the sunsetting process; both
MySpace and Vine were publicly significant social platforms catering to a broad audience. Each had
ties to major corporations and held profit-oriented goals. Indeed, these are disparate sites–in era and
in scope–but in concert they reveal how employee values inflect the sunsetting process.
While I focus on these two platforms because of greater clarity around the handling of user
data during a process of shutdown, I do not solely address them by focusing on the chronological
end of these platforms, for instance, the very last week that they existed. As this project emphasizes,
sunsetting is not an isolated moment. It is not constituted by the pulling a plug on a platform’s
servers, or the final board-room decision to shut down a platform. Conversations about the right
path for a platform, that platforms possible failure or success over time, and the grounds for data
permanence or ephemerality take place throughout a platform’s existence. Therefore, while I often
bring attention to later stages of the sunsetting process where decisions about user content are
especially salient, I draw from all eras of a platform’s tenure to parse how values are expressed and
acted upon.
From Value to Values in Sunsetting
To take a step back: what are values – and why look at them in the platform sunsetting
process? In the preceding chapter, I traced valuation in platform decline– a sociotechnical process in
which user data and other platform components were assigned qualities and value by market actors

104
(Greeson, 2020), including platform executives and venture capital investors. Scholars have dually
argued that valuation shapes and is shaped by what might be called ‘social’ values: those collectively
or individually held core beliefs (Shilton, Koepfler & Fleischmann, 2014) situated in a cultural milieu
and structured by human interests and desires within that milieu (Friedman, Kahn & Borning, 2006).
Social values inform valuation in that collective or individual ‘core beliefs’ undergird the process of
assigning value; that is, value and qualities are not inherent in goods and artifacts but are socially
situated. At the same time, processes of valuation inform social values because social values are not
ossified, stabilized, and static entities, but rather come into being through valuation processes over
time (Houston et al, 2016).
This relationship between values and valuation is an especially important one in platform
sunsetting. As existing user data hosted by platforms shifted from an asset to a burden, as I
described in the previous chapter, employees’ personal values became ever more important in
determining how the sunsetting process unfolded, as I will show in this chapter. Secondly, platform
studies scholars like Van Dijck, Poell and De Waal (2018) have argued that values are what are at
stake in researching platform. Because corporations based in Silicon Valley have created a cultural
environment in which economic value is conflated with public value, this conflation requires a
process of undoing by which societies emphasize social, instead of purely individual or corporate
economic, valorization. Finally, conceiving of user content as moving into the category of ‘waste,’ of
going through the sociotechnical process of being emptied of value, as I described in Chapter 2, has
a particularly salient relationship to social values. Cultural scholars Gay Hawkins and Stephen
Muecke have argued that waste, when it escapes a “perfectly circular model of production and
consumption” (2003, p. x), plays a crucial role in reorganizing (or requiring the reorganization of)
social values. They use recycling as an example of this phenomenon. For instance, in the US, when
goods categorized as waste–milk cartons, yogurt tops, soda bottles–became subject to modern

105
recycling practices in the 1960s, economic assessments wondered if all this sorting, cleaning, and
reprocessing was worth it—whether it had utilitarian value. Yet at the same time, they assert, the
recycling process helped shape a domestic and ethical commitment to environmental concern–
values which, Hawkins and Muecke note, help produce humanistic if not utilitarian value.
Like the artifacts of recycling, user content does not subscribe to a “perfectly circular model
of production and consumption.” Its path is not necessarily to be produced and ‘consumed,’ used
for some defined period of time, then discarded and brought to the trash dump. Instead, user
content seemingly void of economic value is subject to non-circular regimes of valuation: there is
always the possibility that user data could regain economic value through combination with related
consumer data (Spurgeon, 2008); old user content is routinely remixed, remade and recombined
through cultural practices like meme making (Shifman, 2014)—which of course may be picked up
by advertisers looking to cash in on cultural zeitgeist (Kirby, 2013); and importantly, user content
may hold personal, cultural and historical value outside of market relations. How then are social
values organized and reorganized around these valuation dynamics during sunsetting?
Values in Design, Values in Breakdown
Values have rightly been key considerations in research that addresses the design and use of
sociotechnical systems, especially in human computer interaction (HCI) and science and technology
studies (STS). Lara Houston and co-authors (2016) have described a dialectical relationship between
values scholarship in these two fields. As they note, HCI scholarship has called attention to values in
the process of designing sociotechnical system, while scholarship from STS has described how
values can be embedded in infrastructures, and therefore have consequences for the distribution of
knowledge and power within social orders. HCI, in turn, has brought these observations back to the
design process through methods that take systematic approaches to values in both analysis and
design. These systematic approaches have resulted in methods like value sensitive design (VSD)

106
(Friedman, Kahn and Borning, 2006), an approach that accounts for human values throughout the
design process via the identification of stakeholders and the articulation of specific values.
At the same time, value sensitive design and similar approaches have been critiqued as
offering an understanding of values that is static, functioning unchanged through time, existing
outside of action rather than coming into being through it (Jafarinaimi, Nathan & Hargraves, 2015).
Literature on repair and maintenance work––like the kind explored by Houston et al (2016) in their
discussion of values in fixer collectives in the US and mobile phone repair sites in Bangladesh and
Uganda––has been helpful in showing how values are associated with process and action. Like the
framework offered by “broken world thinking” (Jackson, 2014) and other literature on repair and
maintenance (Orr, 1996; Russell & Vinsel, 2020), these studies have noted how “sites of repair
constitute different processes and sites of value than those surfaced in better known instances of
design and use in HCI” (2016). Literature looking beyond moments of innovation, then, has
expanded not only notions of what value is, but also how it comes into being and where in
sociotechnical processes values can be expressed.
While approaches like value sensitive design have been critiqued, approaches like these have
been crucial in showing the ways that the values brought to the design of sociotechnical systems are
consequential. This idea has, in contrast, not been emphasized in the important but limited work on
values in maintenance, repair, and breakdown. This chapter therefore bridges this gap through
sunsetting, asking: which values are expressed through these processes? In what ways are they
consequential? How do these values shape what remains of these platforms?
What’s at Stake When We Talk About Platforms and Values
As noted earlier, scholars like Van Dijck, Poell and De Waal (2018) have argued that Silicon
Valley-based platforms have created a sociopolitical environment––what they call “the platform
society”––wherein economic value is conflated with public value. Indeed, the authors, in their wide-

107
ranging treatment of major platforms like Facebook and Google, have noted that values are central
considerations in understanding platforms’ power: “which values are at stake […] are central in
disputes concerning the creation of public value in the platform society” (p. 25). That is, what the
authors call “public values” ––or the values of a society, typically structured around a common good
(Moore, 1995)––have interfaced with and are often challenged by the value systems presented by
platforms. (These value systems might include oppositions like democratic transparency versus
corporate opacity, or the differences in public accountability between governments and platform
organizations.) Ultimately arguing that the health of a democratic society is at stake when
considering platforms and values, they say that platform research should seek to understand how the
values of a society interface with values that are “built into” the architecture of platforms.
What values have Silicon Valley-based platforms then been shaped by? While nothing exists
outside of social value systems, social media platforms have long been associated with particularly
idealistic value-laden discourses. These discourses did not arise without precedent when the first
platform went online. In From Counterculture to Cyberculture (2006), cultural historian of technology
Fred Turner has shown how the idealistic values of the countercultural revolution in the 1960s US
were transmuted into idealistic thinking about computing and the Internet in the 1990s, framing
these devices as tools for social liberation and alternative community making. Texts like Electronic
Frontier Foundation founding member and internet activist John Perry Barlow’s 1996 “Declaration
of the Independence of Cyberspace” encapsulated this ethos, proffering the idea that internet
technology offered a means by which to supersede the normative value systems of the US and the
US government. Speaking to this normative society from the perspective of the cyber-society,
Barlow opined:
Your legal concepts of property, expression, identity, movement, and context do not apply
to us. They are all based on matter, and there is no matter here.

108
Our identities have no bodies, so, unlike you, we cannot obtain order by physical coercion.
We believe that from ethics, enlightened self-interest, and the commonweal, our governance
will emerge. Our identities may be distributed across many of your jurisdictions. The only
law that all our constituent cultures would generally recognize is the Golden Rule.

These discourses were interrupted by the dot-com crash of 1999-2001, the precipitous downturn of
the internet economy wherein innumerable companies folded. In the wake of the crash, narratives
emerged from Silicon Valley techno-cultures that described a fundamental shift in the capabilities
and potential of the web. Encapsulated by the notion of a ‘Web 2.0,’ these discourses attempted to
communicate to both investors and the public that this new cache of web-based startups had learned
from the failures of ‘Web 1.0’ (Ankerson, 2018). While attempting to distance itself from an earlier
era of the internet, narratives of Web 2.0 continued the idealistic rhetoric of these earlier
technological moments. These narratives put forth a vision of a web that would consist of platforms,
or programmable digital architectures, which could realize the communitarian potential of the web
through their ability to harness “collective intelligence.” Tim O’Reilly (2005), the Silicon Valley
publisher and Web 2.0’s unofficial spokesman, would describe this harnessing of collective
intelligence of the creation of the web as a “global brain,” a disembodied communitarian vision not
entirely unlike the one espoused by Barlow before.
As Zimmer (2008, para. 1) summarized, Web 2.0 discourse emphasized values of self-
expression and information democratization. As he noted, it communicated that:
…everyone can and should use new Internet technologies to organize and share
information, to interact within communities, and to express oneself. It promises to empower
creativity, to democratize media production, and to celebrate the individual while also
relishing the power of collaboration and social networks.
These values helped frame platforms and their role in society to a broader public (Gillespie, 2010).
Indeed, the very choice of the term ‘platforms,’ and its rhetorical position by platform companies,
helped further entrench these associated values. The multifaceted definition of ‘platform,’ whether
as a raised surface, a political agenda, or a space from which to speak, has helped platform

109
companies orient themselves alongside an imaginary of accessibility, egalitarianism, neutrality, and
openness (ibid). Critical work on platforms has ventured to show that these values function publicly
and discursively if not practically (e.g. Gillespie, 2010; 2018; Noble, 2018; Roberts, 2019; Van Dijck,
Poell & De Waal, 2018). That is, platforms’ association with these seemingly democratic values is
strategic, helping companies at once elide regulation while appearing indispensable for fostering
speech and community. What this research presents is the need to look at the relationship between
platforms and values in public–or how platform companies wield particular values to frame their
role in society–and in practice, or how values are expressed through production-level platform
activity, and how employee and organizational values come to shape this activity.
One way to understand the relationship between values and platforms in practice is by
focusing on values at the production level, meaning the values expressed by those producing a
sociotechnical system. As I noted in the Introduction to the dissertation, focusing on production
level values in the context of sunsetting is particularly important because of the outsize power that
platform companies hold in determining what happens to user-generated content. Platform scholars
have argued that platform practices–for instance, content moderation–are shaped by the values that
platform producers–for instance, employees like product managers and developers–possess
(Gillespie, 2018). In so doing, production level values come to influence the public discourse and
social interaction that increasingly happens through platforms (ibid).
Critical examinations of these perspectives have shown the value systems enacted through
platform practices indeed reflect the value systems held by company employees (Gillespie, 2018;
Gipson, Corry & Noble, 2020; Noble, 2018). In the case of major US-based platforms, and
computing cultures in general, these are often perspectives of a limited and privileged group of well
renumerated, white, male, educated and liberal or libertarian employees. Normative values related to

110
this subject position therefore shape a platform’s orientation to society (e.g. Corry, 2021), to its
users, and to public discourse.
Seeing Values
Understanding that production level values shape a platform system’s interaction with
society, how can social values be seen empirically? Understanding the values that employees brought
to the sunsetting process was not as easy as asking participants what their values were. Moreover,
because values are not static entities, as scholarship at the intersection of HCI and STS have argued,
identifying values was not as straightforward as simply looking for the expression of commonly held
values, like loyalty, or compassion, or integrity. Rather, my approach to identifying values in the
sunsetting process was derived from specific but overlapping definitions of what values are and
scholarly descriptions of how values are expressed. Friedman, Kahn and Borning (2006), for
instance, have offered a shorthand definition of values as “what a person or group of people
consider important in life” (p. 349). In this sense, values emerged in conversations and in thematic
analysis as aspects in the sunsetting process that took on greater importance or were identified as
priorities. Values therefore commonly emerged in detailed conversations about sociotechnical
processes, wherein employees offered descriptions of why one path was taken and not another.
Values can also be seen through what Antonio Gramsci (1948/2011) has called “common
sense”–a disparate set of beliefs held commonly within a group, and a perspective on what is ‘good’
that is inevitably influenced by institutions and social hierarchies (Crehan, 2016). To put it more
simply, what is believed to be the correct or logical step–the “common sense” solution–is one way
of seeing values in a sociotechnical process. These values are also made visible through visions of
what the “right way” to do something looks like, even if that perceived right way was never taken.
Finally, values were often highlighted when employees discussed moments where their
perspective or approach clashed with broader organizational perspectives. Moments like this, as will

111
be seen in this chapter, sometimes even engendered a participant to, without prompting, explicitly
use the term “values” to describe their perspective on sunsetting. These lenses–understanding that
values become visible through priorities, through notions of common sense, and through
opposition–help parse what values meant and why they mattered in the sunsetting process. In what
follows, I draw on these lenses to highlight production level values in the process of sunsetting
MySpace and Vine.
MySpace: The Thrill of Creative Destruction
As I described in the preceding chapter, organizational priorities shift swiftly and
significantly once a platform is recognized as no longer growing user engagement or attracting
cultural interest in the way it was previously. That is, things change when a platform no longer tracks
upward on the hockey stick of growth. In this section, I pick up the story of MySpace where it left
off in the preceding chapter: largely considered as culturally irrelevant within the United States, and
then purchased by Specific Media in 2011. I focus on this era, the sunsetting of what employees
called the “old MySpace,” and the process of relaunching what was called the “new MySpace”
between 2011 and 2014. In this transition, MySpace employees had to grapple with what to do with
user data from the “old MySpace” that no longer maintained the economic and cultural value that it
once had. Some of these employees had worked for MySpace during its heyday, some were hired to
relaunch MySpace, and some transitioned from working on Specific’s other verticals to work on the
acquired platform. Most of this section focuses on the values that these employees brought to this
process. At the same time, and as I have argued throughout this project, this era is not easily
separated from the ones that preceded it. Because sunsetting is not just the pulling of the plug, I also
highlight other moments from earlier in MySpace’s tenure that speak to employee and organizational
attitudes toward data management over time.

112
“Churn and Burn” – 1999-2003
The philosophy that undergirded MySpace’s development at eUniverse was one described to
me as “churn and burn.”
16
This meant that the eUniverse approach was to come up with a new idea,
build it quickly, and see if it stuck–meaning that it attracted user interest and resulted in profit
rapidly. This is how the company had generated the sites and services described in the last chapter,
from the selling of e-scooters and anthrax test kits and CDs online, to the sites featuring
inspirational quotes and e-cards. Under the philosophy of “churn and burn,” if a site did not quickly
result in profit, then the company reasoned that they had not invested too much time, energy, or
resources, and could move on without incident. “Churn and burn” was a philosophy that resulted in
“never [having] a long-term structure in mind,” as one interviewee noted, for MySpace or any of the
other eUniverse products and services.
In other words, MySpace was developed in an organizational context that prized and
prioritized immediate success–and immediate failure. This “fail fast” ethos is one that is well known
and much practiced in American techno-cultures (Draper, 2017), a philosophy that proponents
believe combats staleness and promotes innovation, with rapid failures serving as pretexts for more
successful products to follow (Corry, 2021; Lipartito, 2003). Fail fast, as it was practiced at
eUniverse, however, appeared to not necessarily be the desire to iterate upon a particular product as
it was to move on quickly to a different focus, and see what else could drive the company’s bottom
line. As I discuss in the next chapter, this ‘build quick, fail fast’ attitude informed the technical
construction of MySpace, rendering it an unwieldy system that was difficult to maintain and,
ultimately, difficult to take apart.

16
A more common definition of churn and burn is a union-busting tactic wherein large corporations cycle through
employees until they’ve hired a majority that are anti-union (Chung, 2011).

113

The Social Network Contract – 2003-2009
2007-2009 represented MySpace’s inflection point, a span when the site was operating at
peak usage, then was surpassed by Facebook in numbers of users, and finally began to take a
precipitous downturn in user engagement. I spoke to employees who worked during this period, and
were familiar with the finances of the company, especially how the platform budgeted, leased, and
projected for the computational power and storage infrastructure that the platform needed over
time. Quarter after quarter, employees would project what were thought to be wildly high
calculations for the technical infrastructure the platform required to support its growing userbase:
and still, each time, the infrastructure that was needed surpassed these projections.
At least until 2009. MySpace’s drop in usage over the course of that year was precipitous, the
team responsible for making these projects were some of the earliest employees to deal with the
infrastructural and technical ramifications of MySpace’s downturn. The platform was, as one
employee phrased it, “working until it wasn’t working.” In response to this downturn, MySpace
engaged in drastic infrastructural economizations, requiring the re-negotiation of years-long leases
that they had made with data centers, along with other tactics meant to reduce overhead costs. As
one employee familiar with these processes described:
We [scaled] back, cut down budgets like crazy. Certain commitments that were made we
ended or didn’t continue or negotiated. You can pay less for powers and space, and idle the
servers if you’re not using them.

Given the drastic measures the company took to economize, I asked this interviewee if MySpace had
ever considered cutting storage infrastructure to cut cost. The idea was clearly taboo:
If you had a profile and put up pictures of yourself and songs, you have to store those
assets so whenever people come to look at that they see those things. You couldn’t delete
that, or that was not our policy. That would be weird.

114
By remarking that deleting user data to economize on storage costs would be “weird,” this comment
highlights a commonsense value when it comes to data management on social media platforms. That
is, the maintenance of user data is prioritized–at least when a platform is still perceived as
maintaining some value. This could be called a social network contract: the implicit agreement between
users and platforms that, so long as users place their content there, it will be available for retrieval.
This contract highlights implicit user expectations and implicit platform responsibilities toward the
maintenance of user content: in doing so, it shows the ways in which platforms embody aspects of
infrastructure discussed in Chapter 1, as they invisibly support practices of recall, recall for memory
among them.
From Old to New MySpace – 2011-2014
In the United States, MySpace had largely faded as a preferred profile-based social media
platform by 2010. In 2011, Specific Media–an internet advertising company specializing in placing
ads on remnant online ad space–purchased MySpace for $35 million dollars, with the idea of
relaunching the network as a music artist portal. The vision for it, as it was told to me, was
somewhere between a Spotify, the popular music streaming app, and a social network. In 2013,
MySpace completely relaunched, transferring existing user accounts to the new platform but deleting
significant portions of these profiles along the way. Data that was removed included blogs, private
messages, videos, posts, and comments on those posts (Lomas, 2013); it also included the fan bases
that were accrued on music artist accounts (Mulligan, 2013). While users had been notified that their
profiles would be transferred, many were surprised at the content removal (Lomas, 2013).
During this transition, MySpace and Specific Media negotiated a central question: was it
possible that a social media platform could succeed after it had declined in economic might and
cultural relevance? How should one reconcile the past and the future of the site? And what should
be done with the user content, user data, and user communities that continued to exist in varying

115
forms on what employees took to calling the ‘old MySpace’? As the MySpace/Specific team began
designing what they called both publicly and internally the ‘new MySpace,’ they considered both the
economic value, their perceptions of user communities, and their perceptions of themselves in the
process.
Considered Values. Nassim Jafarinaimi and colleagues (2015) have argued that values
shape design processes when producers “encounter problems of how to serve the many demands of
human life and living in particular and changing circumstance” (p. 96). They argue that values can
therefore be understood as hypotheses: helping producers navigate possible design decisions in varying
contexts, which lead toward particular “valued ends.” Understood in this way, it is important to
consider values not exclusively through ends, but also how values shaped the entire landscape of
consideration in the design process–or, as Jafarinami et al. phrase it, by also considering valued
means. This perspective resonates with work from the history of technology and STS, which has
shown the utility in considering counterfactuals, or those paths that were not taken in the design of
an artifact, to reveal the social construction of that technological artifact (Brown, 2014; Cowan,
1985).
This perspective is necessary for understanding how MySpace employees grappled with
values in the process of sunsetting the ‘old’ MySpace. While the valued ends of MySpace’s sunsetting
process in 2013 may be clear–the data deletion that was previously described–these ends do not
straightforwardly reflect the broader universe of considerations that were brought to the handling of
user data. Indeed, as employees worked on building out the new platform and making decisions
about how to handle existing user content, they surfaced multiple and sometimes competing values
in the process.
Julian, whose quote began this chapter, was a senior content writer during MySpace’s
transition under Specific Media. Joining MySpace to help steer its remake, he was tasked with

116
explaining to the public what this ‘new’ MySpace would be and what its relation to the old MySpace
was, as he wrote copy for TV ads, for emails to users, and for the site itself. As such, he was
involved in conversations over what should be done about existing user content, and how that
decision would be explained to users. Julian explained a host of considerations–a host of values–
through which MySpace grappled with how or if to maintain user content. He began by noting that
these decisions were not taken lightly:
I remember the conversations around people’s content being really, really heavy. Because,
for the majority of people who stopped using MySpace, their page was there and they never
touched it again, and they never went back. There were some people who would maybe look
at it once a year. But there were a lot of edge cases that I remember being talked about that
put you in this really philosophical place about, what do we do?

The first consideration that steered their actions was legal in nature:
For instance, I think there were cases where MySpace pages were involved in trials. I can't
remember if they were specifically murder trials, or disappearance [trials]. There was active
police work going around some of these pages.

Yet they also considered how people might use these pages for more intangible purposes like
personal memory and memorialization:
There was also talk about people who had passed away and these pages were memorials to
people who’d passed away. There might still be people going there. There was also a small
but not insignificant group of users, that just the nostalgia of it was something that they
enjoyed. They liked going back to high school, and seeing that stuff, or they liked going back
to middle school and seeing those things.

Finally, Julian discussed the novelty of the issue of long-term content management:
The legal stuff was above my pay grade […] There are laws out there about people's data and
protecting it, and ensuring that they have access to it in some way or another. But [for the
other considerations], this was one of the first times, at least in my digital life, that I
remember having to even think about this stuff.

117
As Julian summarizes in this extended quote, MySpace employees did not just rely on a singular
value to guide their decision making. Rather, this process elicited competing forms of valuation and
competing forms of social values that were in turn drawn on to navigate this context.
Above all, legal compliance was considered the most important and the most known
parameter in the maintenance of user content. At the same time, this content represented a relatively
small and well-known collection to maintain; it was moreover a collection that did not need to be
maintained publicly. Employees also weighed the importance of the site for memory, and identified
memorial reflection by family and friends as a meaningful example of continued use. However, as
Julian attested, these profiles were largely considered to be “edge cases,” because these pages were
visited infrequently. Finally, this scenario shows how employee values were drawn on in a moment
that seemed to offer little precedent, and few known rules. In the rest of this chapter, I’ll discuss the
other values that were elicited in this process, and how they ultimately shaped the “valued ends” that
were pursued.
Creative Destruction. The “heavy” considerations that Julian outlined around how to
handle user data were largely overshadowed in my conversations by excitement at the prospect of
building the new MySpace. In contrast to the discomfort of discussing sunsetting, the new MySpace
elicited enthusiasm as employees recalled their time attempting a ‘moonshot’ and working in a
creative and ambitious environment. As David, who played a senior leadership role on the content
production team, described to me:
That was really where the electricity in the room happens, when you have a blank
whiteboard and you have funding and you have just your ideas. […] At lunch you're sitting at
a table […] and you're talking about some of the wildest shit you can think of. You're going
for a walk around the neighborhood with someone on your team and they have this great
idea. There’s a lot of electricity to that before you launch.

The “electricity” that David describes was based in part on having a “blank whiteboard,” what he
later described in our conversation as “carte blanche,” an open and unencumbered space to create

118
the platform anew. Yet, as explanations like Julian’s have shown, building the new MySpace did not
in fact represent unencumbered social and technical space. Rather, employees necessarily, if
begrudgingly, confronted the old MySpace that was in the process of being sunset. While employees
valued the potential for creative novelty, the fact that they also had these “heavy considerations”
required that employees navigate varying priorities to decide what to do with the old MySpace data.
I want to suggest that employees were able to navigate these two competing values–on the
one hand, a latent sense of responsibility to the user community that had been fostered on the old
MySpace, and on the other, a desire for unencumbered creative novelty–by framing this process of
sunsetting as one of “creative destruction.” Creative destruction is a phrase associated with the 20
th

century Austrian economist Joseph Schumpeter and his theories on innovation, which signaled a
process whereby new technologies, products, modes of production, and modes of distribution make
existing methods obsolete (Reier, 2000; Vinsel, 2014a; 2014b). This requires organizations to adapt
to these new modes, innovate themselves, or fail; in other words, innovate or be destroyed.
However inaccurate Schumpeter’s models are in describing real economic processes (see Lepore,
2014), his theories were adopted widely as gospel, especially so in Silicon Valley and related techno-
cultures (Vinsel, 2014a; 2014b; Vinsel & Russel, 2020). Adherents to this philosophy valorized
entrepreneurs, bemoaned market stagnancy, and lauded the boom-and-bust market changes that
were a result of innovation processes, that “perennial gale of creative destruction” as Schumpeter
himself had put it (1942/1962, p. 82-83). MySpace employees, in other words, could reconcile the
deletion of the content from the old MySpace through a philosophy of innovation, arguing that the
“creative destruction” of the old platform was necessary, natural, and beneficial.
As MySpace transitioned between the old and new platform, ‘creative destruction’ was at one
point literally enacted. Julian recounted the ways that MySpace considered introducing the new
platform to its existing users. One of the ideas they pursued was sending a video of an elaborate

119
pyrotechnic stunt to its former users, one that was a metaphor for the destruction and death of the
old platform. As he recounted, prior to their purchase by Specific Media, an expensive wooden
bench in the shape of a MySpace logo welcomed visitors to their building (Figure 5).

Figure 5. Wooden bench in the lobby of MySpace's office (“MySpace HQ,” 2011).

Because MySpace was changing their logo, they decided to scrap the bench: drive it out to the
nearby desert, douse it in gasoline, set it on fire, and film it. Eventually shown at a companywide
meeting, Julian noted how the film of the burning logo served as a motivating call to employees to
embrace innovation through the new MySpace: “Hey, we’re doing something new. Everybody get
on board. Get excited. It’s going to be awesome. It’s going to be wild. It’s going to be fun.” Part of
that, of course, involved the shedding, or the “creative destruction,” of the old. As Julian

120
summarized, the burning of the logo was “just this crazy amazing visual metaphor of what was
about to happen.”
Desired Community
17
. Not everyone on the MySpace staff responded well to this ethos, to
“get on board” and “get excited” about building the new platform by way of destroying the old.
Rather, employees described to me a tension-filled process wherein new employees were
emboldened by these visions while those who had worked for MySpace before the transition under
Specific were more reserved in their enthusiasm for the wholesale destruction of what had come
before. This tension was expressed especially in organizational conversations about user
communities, conversations that typically juxtaposed those still active on the legacy platform with
the users that were envisioned for the new platform. Gavin, a product manager hired under Specific
who played a significant role in steering MySpace’s transition, described how employees who had
worked for MySpace prior to Specific often “had a real allegiance to the existing [user] community.”
But Gavin, like other new hires, “felt like the existing community were kind of like … I
don't want to say lesser humans, but lesser value.” Gavin rationalized this statement by arguing that
the existing user community “had a different interaction style. They were comic book fans, anime
fans that were talking to each other. […] They would have cartoon avatars. They were people that
were inauthentic, or they weren’t themselves.” This style of interaction––this style of community––
went against what he believed to be a governing principle of successful social media platforms,
wherein platform users should be their “real” selves:

17
Community as a term is “notoriously difficult to define,” (Dunbar-Hester, 2014, p. 212), acting as “a highly
moralized and politicized keyword” (Avance, 2016, p. 63). Like the discursive approach I take to the term
platform and content throughout this project, community in this project–as it is typically used by
interviewees–can be thought of as the users of a particular platform (e.g. a ‘user community), connected
through the sharing of a common communication network. As Raymond Williams (1976, p. 76; cited in
Dunbar-Hester, 2014) has also noted, community is often a “warmly persuasive” term to describe
relationships and was used with this inflection at times by my interviewees–in this chapter, this was the case
more with Vine than with MySpace.

121
The lessons of social media have always been that you need authenticity in order to have a
soluble community. If you have a group that can interact on the basis of shared ideology,
that’s something. But it’s not enough, right? Then you just have fan clubs. We wanted to get
people who were themselves, […] people that are authentic creatives that want to share their
photography, their music, their artwork, whatever it was.

Gavin draws on visions of what the “right” community for the new MySpace was in this quote,
gesturing at desired visions of what authenticity, creative production, and meaningful interaction
meant to himself and other new hires. Visions of a desired and ‘authentic’ community for the new
platform steered design decisions as they built the new platform. As Gavin described, the new
platform’s design was oriented around questions like “What are the features that are going to attract
the types of community members that we want?” Questions like these highlight the values that were
brought to the construction of new features––but they also implicitly point toward the values that
were brought to the process of sunsetting. In other words, as visions and values of a desired
community informed the active design of the new platform through new components and features,
these visions and values also shaped the sunsetting process, shaped what would happen to existing
user content and any existing user communities.
What Would I Want? Mapping Employee Values to User Values. In addition to visions
of desired community, MySpace employees brought their personal beliefs around content
maintenance to ideas of what the right thing to do with existing user data was. Beyond simply
expressing these beliefs, employees mobilized personal beliefs in the sunsetting process by mapping
them to what they presumed to be user beliefs–a implicit design process that STS scholar Nelly
Oudshoorn and colleagues have dubbed “I-methodology” (Oudshoorn et al, 2004). Employees then
used this alignment as rationale for decisions–especially decisions around content deletion–in the
sunsetting process.
Gavin, for instance, drew on his own perceptions of personal privacy online to frame
MySpace’s history of content removal. “This goes back to my request for a pseudonym,” he began,

122
referring to his preference that I obscure his real name when writing about anything from our
interview:
I bet there's a lot of people that are happy. There are some things that are really difficult to remove
from the internet, and anybody that really works in this space I think is a lot more cautious
around what goes out about them. I would submit to you, that some things are better lost.
[…] Sometimes you burn the shoe box, you know what I mean?

For Gavin, his own trepidation toward personal information circulating online, and the unwelcome
permanence of this information, informed his belief that users were likely happy rather than upset
that their content had been removed. Alluding to his own experience working on internet
publications and platforms, Gavin attests to the notion that content loss on platforms is a unique
opportunity to be rid of the baggage of the past. Rather than a disappointment, data deletion was
like an agentic burning of the “shoebox,” that traditional container for storing personal reminders of
the past.
Lydia, who was part of the creative content team during the transition, felt similarly. She
begrudgingly recalled how some users had found her contact information and reached out to her
after MySpace transitioned to the new platform and removed user data. They had wanted to get their
content back. “I remember psycho people […] being like, ‘Excuse me, I can't find...’ I'm like, what
the fuck?” She compared it to the nonchalant way that she had approached a platform transition as a
Facebook user:
I remember a couple of years ago Facebook was like, “Backup your [content] now because
we're changing it.” And it's like, okay, great––I probably shouldn't have thousands of photos
of me [partying] when I’m 21 years old on the internet. I don't really need to keep those. So
if they go into the ether, too bad, you know?

While Lydia was faced with the reality that some users would still want their content, she drew
attention to what she believed to be appropriate user attitude toward platform content by drawing

123
on herself as an example. Like Gavin, Lydia framed content loss as either beneficial for users,
helping with informational management about one’s past, or simply as banal.
Ambivalence and Irony. Lydia, however, was not entirely unsentimental for digital content
she had once maintained. Instead, like several other employees at MySpace that I discuss in this
section, she expressed an ambivalent, conflicting and ironic attitude toward content loss. Later in
our conversation, as she was explaining the new MySpace by comparing it to the music streaming
platform Spotify, she recounted why she had started using cloud-based music services like Spotify in
the first place: it was because cloud-based platforms did not risk the loss of accumulated content in
the ways that physical devices like CDs, computers or mp3 players did. “I've been paying for Spotify
since it came out,” she noted, “because I've gotten all my CDs […] stolen and my computer stolen a
million times. I lost all that beautiful music that I had downloaded.”
Lydia’s ambivalence toward content loss on MySpace and other social media platforms
contrasted her frustration at the loss of music that she had personally invested in and stored on
various devices. Much in the way Lydia did here, MySpace employees drew distinctions between the
type of cultural production, informational behavior, and content that was considered as valuable. In
this sense, employees at once mapped personal values to user values while separately maligning the
loss of content from other contexts.
David, a creative content producer, also worked on transitioning the platform from the
legacy MySpace to the new iteration, and, like Lydia, was ambivalent about the loss of content on
platforms. Acknowledging the regularity of content loss, he noted that the data loss on platforms
paled in comparison to the loss of music and culture-focused blogs he used to frequent:
I think we who have grown up on the internet think that everything should be forever, and
accessible at all times. In the last five, six, seven years you've seen that's not the case. All
these old blogs that I used to read don't exist anymore. You can't find them, even on the
Internet Archive––and that's a shame. I'm much more sad about that stuff than platforms.

124
Yet like Lydia, David later diverged from this opinion when he referenced his fervent search for
content he had produced for MySpace. Prior to the server loss in 2019, MySpace had apparently
deleted some of the content produced as part of its relaunch that was hosted on the platform. It was
content that David, who spearheaded many of those creative projects, was proud of. Blindsided by
the loss, David has contacted each new owner of the platform as it has changed hands––from
Specific to Viant to TIME Incorporated to Meredith Corporation––to see if they had a hard drive
somewhere with his content on it.
Michael, a content producer, worked closely with David on this creative team to generate
campaigns that would bring a certain underground cool to the site. Trepidatious at first in our
conversation, he eventually grew candid as we discussed the ins and outs of his time working for the
platform. Like in Michael’s case, employees spoke at length about the bad decisions they thought
executives had made, the infighting at upper levels of management, and the ego and greed they
believed drove some company-level decisions. These moments often went by without participants
emphasizing their pseudonymity: surprisingly, it was when I broached content loss that concerns
over pseudonymity came to the fore.
We were at the tail end of our conversation when I asked Michael what he had thought
about the loss of content caused by the botched server migration in 2019. In what seemed to be a
curious contrast, Michael emphasized that, because of his attitudes toward the maintenance of
content, he strongly requested pseudonymity lest it risk his future employability in the digital content
space, requesting that I not put some of his “language out in public” to “make sure that future
employers would still hire me.”
“I got a real mercenary attitude towards this stuff,” he said. Having worked in a variety of
cultural industries, from music to magazines, he had seen his favorite underground publications
shutter suddenly––especially during the global financial crisis of 2008. The swift decimation of

125
longstanding cultural forums that he had witnessed colored his perception of the loss of platform
content. “For me, losing that content [on MySpace], it's disappointing, but I'm also really just jaded
about it all. It's just like, ‘Of course.” There was more than a tinge of anger and disappointment. “Of
course we don't have it.”
As in this exchange, throughout our conversation Michael toed the line between a devil-may-
care attitude toward content loss and frustration over the prevailing dynamics of cultural access and
preservation. At one point he pointed out that his “que sera, sera, what's in the past is in the past, and
keep moving type attitude, is not the same as everybody else. I don't have a lot of personal
snapshots. I'm not an archivist like that.” (Unlike others I interviewed, he was careful to note that
his indifference toward content loss was particular to himself and not representative of others.)
Despite the blasé approach to content preservation seemingly expressed here, he later vehemently
addressed how issues around informational access influence the entire landscape of cultural
production, and the history that can be told about these cultures, appearing to contradict his earlier
apathy toward what gets saved and what doesn’t get saved. As he noted: “There's all this discussion
about what's lost in any history when everything is digitized. The catalog gets so narrowed based on
what's available and who has rights to what. […] And that's not just music, it’s movies, it’s
literature.”
Interviews like these––that contained contradictory, ambivalent or ironic statements––
surfaced tensions in how MySpace employees navigated the sunsetting of the old platform. On the
one hand, these conversations showed how employees used personal values to navigate and explain
organizational decisions around user content, in line with observations from Jafarinaimi and
colleagues that values act as “hypotheses” (2015). At the same time, these contradictory exchanges
point toward an even more pronounced notion that values in practice are non-static entities: not
bounded beliefs that can be named, wielded, and straightforwardly embedded in the design of a

126
system. As can be seen here, this is in part because values can be internally contradictory or may
only become salient in particular situations.
The question remains over why MySpace employees displayed such internally contradictory
approaches to user content, why they could go from attesting to the triviality of content loss on
platforms to lamenting the loss of their own digital data. For one, the content that employees felt
was valuable was not necessarily hosted by online social media platforms; Lydia, for instance,
discussed content hosted on digital storage devices, and Michael referenced the loss of traditional
print media. Perhaps employees felt ambivalent about the types of content and abundance of
content that flow through social media platforms when compared with these other types of cultural
production they identified as more valuable. Moreover, employees may separate the loss of
personally valued content with the issue of broader content loss on platforms; indeed, when
interviewees mentioned how data that had personal value was lost, this was generally at a different
juncture in our conversation than the explicit discussion of MySpace’s handling of user data during
sunsetting. Finally, because employees exhibited discomfort when discussing decisions around user
data deletion in the first place, they may have felt the need to justify this decision so that it does not
appear unethical–for instance, by arguing that users would want content loss.
Importantly, the values surfaced in the sunsetting of the legacy MySpace platform––from the
early ethos of churn and burn and the social contract of persistence, to later values of desired
community and creative destruction–show how sunsetting is not an incidental or value-less process,
but instead is shaped by a complex negotiation between organizational, employee, and user beliefs
around what is important and what is right. As I’ll discuss in the next chapter, these values
eventually influenced what remained materially of MySpace after this 2013 closure.

127
Vine: Choosing the ‘Right Way’ to Shut Down a Platform
Vine, the platform in which users recorded six-second, looping videos, launched auspiciously
on January 24, 2013. The first widely popular mobile video application in the US, it was purchased
by Twitter before launching. As it was told to me by Vine employees, this rapid acquiring occurred
in part because at least one of the app’s founders––Rus Yusupov, Dom Hofmann, and Colin Kroll–
–was friends with Twitter CEO Jack Dorsey, in part because Twitter was looking to expand its
capabilities into video, and in part because Twitter had famously missed out on the purchase of
mobile image platform Instagram, which was acquired by Facebook in 2012. Moreover, Vine–with
its short and snappy six second limit–seemed to resonate with Twitter’s text-based character
restrictions.
By October 2013, Vine had 40 million users, making it the fastest-growing app in the world
(Graham, 2013). It became known as a place for rapid and easily memed forms of entertainment,
particularly a brand of frenetic comedy (Roettgers, 2016), while elements from the app’s videos
gained cultural and often linguistic ubiquity (Rosza, 2016). Some popular Vine users became
celebrities with agents and entertainment deals, helping launch an industry centered around social
media ‘creators’ (Friend, 2014).
18
Others flocked to it because of particular communities that used
the app: whether community based on shared identity or community based on specific cultural
interests. Vine, for instance, became known as a place that showcased the work of young Black
people (Calhoun, 2019; Lee, 2017), becoming, as Jazmine Hughes (2016) would write in the New
York Times, “a digital exposition of black achievement.” As employees recounted to me, it also
became a hub for burgeoning international community interested in K-Pop, or pop music

18
Several Vine stars lived and created together at an apartment building in Los Angeles on the corners of
Hollywood Boulevard and Vine Street (Friend, 2014).

128
originating in South Korea, just as it was a destination for American sports fans interested in
following rapid cuts of their favorite teams’ highlights.
However, Vine’s golden era started to fade. Instagram debuted video sharing in 2013,
mimicking Vine’s core feature, and in the ensuing years, the press began to speculate about Vine’s
waning interest from advertisers while rumors swirled of Twitter’s diminished investment in the app.
Looking back from today, industry insiders venture further reasons for Vine’s ultimate downfall: that
Vine had been neglected by its parent company Twitter; or that Vine had neglected its creators, who
were a key driver of traffic for the app; that it had fallen under mismanagement by its founders, who
exited the company before Vine was sunset; or Instagram and Snapchat had mimicked its features
and taken its audience. These were all put forth as root causes as to why Vine shut down, but the
ultimate reason that employees I spoke to seemed to agree upon was that Twitter–which went
public in 2013–needed to trim components of its business that were not obvious money-makers in a
quest to appease shareholders and become profitable.
Vine announced its closure in a Medium blog post in October of 2016. In it, its founders
made a profoundly value-laden statement: they promised that they would shut down the platform
“the right way.” As they wrote:
We value you, your Vines, and are going to do this the right way. You’ll be able to access
and download your Vines. We’ll be keeping the website online because we think it’s
important to still be able to watch all the incredible Vines that have been made.

The particulars of this right way, however, were not immediately obvious. As the discussion of
MySpace earlier in this chapter makes clear, the right way to sunset a platform is subject to a wide
range of value-laden considerations, to contradictions in priorities, and to the presence or absence of
hard and fast rules around data maintenance. In this section, I examine how employees at Vine
grappled with what it meant for a platform to shut down the “right way,” as they navigated personal
and organizational values in the construction of Vine’s afterlife.

129
Anger
The public had found out about Vine’s closure on the same day that most Vine employees
had. The night prior to the announcement, Vine staff received an email asking for all to be present
at a meeting the next morning. Most knew something bad was going to happen: the email was
written in a “happy tone,” as one person described, but its vagueness seemed to signal something
more ominous. Expecting layoffs, employees braced themselves for an announcement that some of
the 80-odd Vine employees were going to be cut. Employees arrived that morning to find out that
layoffs were indeed happening–but they were for everyone. More shocking was the news that, along
with the layoffs, Vine itself would also be sunset.
The leader of the organization–who did not agree to an interview, citing the negative
emotional experience that Vine represented for her–framed the closure as an opportunity to
celebrate the “amazing work” that Vine staff had done together. The sentiment was received poorly
by most employees. As former Vine designer Paul recounted, it did not go over well because
employees didn’t feel like it recognized the acute loss the staff were suddenly experiencing: “it’s like,
well, some people here are actually losing something they cared about.”
HR representatives came around following the announcement to hand out severance
packages, a moment that music editor Austin described as “like getting served at a five-star
restaurant where they bring in all the plates at once,” as HR staff “came in on the wings with stacks
of the folders.” All employees, except for ten, were told to pack up their desks. The remaining
employees had been selected to stay on for three months to help sunset the app. Among them were
a handful of engineers, a creator representative, designer Paul (who chose not to stay for the sunset),
and the current leader of the organization. As part of my interviews, I spoke to employees who were
laid off that day, employees who were there to steer the sunset, and one employee who was there
during the sunset and continued to work on Twitter, and even on remnants of Vine, after it closed.

130
Employees–even the ones who were going to stay another three months–were upset about
the closure. (As one quipped when I asked what happened after the employees had packed up their
things, they “went out and did some drinking.”) In many ways, their anger mirrored the public
outcry that followed Vine’s announcement of closure––as a New York Times’ headline the day of the
sunsetting announcement read, “Vine is Shutting Down, and the Internet Can’t Stand It” (Rogers,
2016). Indeed, Vine employees were upset that Vine was shut down when a sizeable number of
users still went on the platform. They were upset about how they had been laid off. They were upset
about the organization’s perceived mismanagement throughout its tenure. They were upset at the
cumulative decisions of the founders, and the positioning of Vine within the larger umbrella of
Twitter, as Vine had remained largely separate from the main business. And in the wake of Vine’s
sunset, some were upset about how the shutdown had been handled, and what had happened to
Vine’s content in this process.
This is to say that Vine employees had strong and cogent feelings about how sunsetting
happened. Rather than seeing this process as incidental or banal–as some employees at MySpace
did–Vine staff tended to approach sunsetting as an ethical process, one in which personal and
organizational values were carefully considered. I chronicle this affective orientation toward
sunsetting to begin discussing the values that Vine employees brought to the process of closure not
because the anger expressed by employees is itself is a value, but because employee anger was
elicited when values––like internet culture and community that I’ll discuss herein––were perceived
as being violated. In turn, this affective orientation helped mobilize employees to act on these values
as they designed the sunset.
Internet Culture
What became clear in my interviews are the ways that Vine employees had grown up in an
era when “Internet culture” had become a more cogent phrase. (Perhaps unsurprisingly, my

131
participants tended to be younger the more recent the platform they had worked for began.) While
diffuse as a term–technology scholar Jonathan Zittrain (2014) once noted that the term “internet
culture” is much like saying “electricity culture”–internet culture might broadly refer to the ways that
internet-specific forms of cultural production like memes, hashtags, forms of visual production and
humor are used by groups and individuals, typically on social media platforms (Highfield, 2017).
Vine employees often voiced their deep appreciation for internet culture: especially the
original content, viral phenomenon, creators, and communities they associated with this term.
Moreover, they associated the best of internet culture with Vine. Employees referenced the “thriving
culture” of Vine’s users, as [marketing specialist] Caroline put it, or framed Vine as “a place where
content and culture started,” as Paul described. Austin likened Vine’s creative content to art,
assigning it higher cultural status than is routinely given. Referring to how users across the platform
would iterate on a viral theme, whether a dance move or a particular song, he noted that “the format
creates a totally different type of interaction. It’s a truly like modern expression of art.” He went on
to compare the production of this interactive art with what he believed to be the limiting, one-way
nature of broadcast media, a form of communication that did not foster this shared creative culture.
Creation platforms like Vine are a two-way street. I am an artist and I release something, but
then the creator community interprets that and adds onto it and turns it into a meme. […]
Maybe they add a dance to it. Maybe they add a certain creative quirk to it. […] A creator
community takes that song and adds it to this different type of thing, and that applied
meaning really creates a [situation where] one plus one equals three. You have a situation
where it takes on a whole new life and meaning that is greater than what the artist could have
just done on their own.

Vine employees approached sunsetting in a way that reflected their belief that important parts of
popular culture were fostered specifically through social media platforms, and especially so through
Vine. Because of this, employees who steered the sunsetting wanted Vine content to continue to be
accessible to users, even if the networked nature and community interaction itself was removed.

132
As such, Vine employees who were there for the sunset advocated for what they believed to
be “the right way,” a way that preserved Vines in perpetuity, a way that diverged from the initial
organization-led plans for the sunsetting. As creator representative Becky told it, the initial plan was
to keep Vine content around for six months after the shutdown. For that limited period, users
would be able to download their Vines, and Vine videos that had been posted on other platforms,
like Twitter or Facebook, would also continue to play. After those six months, “all of the historical
Vines that were posted on Twitter would never work, and you would just see a frowning face instead
of a video. Anybody who wanted to post a [existing] Vine [on a different platform] in the future
would never be able to.” Employees therefore prioritized the persistence of Vine content because
they believed this content was an important part of popular culture but also because removing this
content would cause gaping holes on other sites and other social media platforms.
Employees contrasted their investment in internet culture with what they understood to be
corporate-level values around profitability. As it was described to me, the larger Twitter organization
wanted an expeditious and most of all a cheap shutdown. Maintaining content takes resources, as I’ll
show in more detail in the next chapter; it requires energy, technology, money, and employee time.
The sunsetting team therefore had to fight, collectively, for the organization to allocate resources for
Vine videos to remain accessible in some form. Drew, who was the primary technical architect of
Vine’s sunsetting and who continues to work at Twitter today, phrased this clash between the Vine
sunsetting team and Twitter executives explicitly in terms of values.
To be as honest as possible, it wasn't that this organization had these ideas, that intrinsically
that they had these values or anything. It was just a very small number of people who
decided that […] we should try to preserve what we can. I can say with no doubt that not
everyone was onboard with this, not everyone appreciated this. We were very much seen as
rogues in pushing this agenda and trying to suggest that we're going to live out past this
specific timeline [of 6 months]. While it would be nice to think [the organization] just
intrinsically had these values, that was definitely not the case. It was just really the
determination of a few individuals.

133

Referring to this group of employees as “rogues” in disrupting Twitter’s sunsetting plan by arguing
for a more permanent archive, Drew evokes the language of media scholar Abigail de Kosnik in her
book of similar name, Rogue Archives (2016). As I’ve examined in more detail in the Introduction, De
Kosnik considers how groups of nonprofessionals, including amateurs, fans, and hackers, have
engaged in digital archiving practices to save the records of online communities that would not
otherwise be represented by traditional memory institutions. By referring to them as “rogues,” De
Kosnik means that these groups took the “initiative to design, found, and run their own cultural
memory institutions without waiting for traditional institutions to set any precedents for online
archiving, [thereby] achieving a degree of democratic inclusion and access for which brick-and-
mortar archives never even aimed” (p. 2). Like the diverging values between corporate-level
executives and Vine employees that Drew describes, De Kosnik’s characterization of these
nonprofessional archival groups as “rogues” highlights how they confronted opposing value
systems, though in this case they were the values of traditional archival institutions. Just as values
spurred these nonprofessionals to engage in diverse forms of digital archiving, so too did these
values for Vine employees–here, reflected through the discourse on the meaningful nature of
internet culture–influence the construction of the site’s afterlife.
Responsibility to Community
Inseparable from Vine employee’s belief in “internet culture” was their enthusiasm for
internet community, or groups who came together online out of shared interests and shared
identities. A sense of responsibility to varying communities pervaded my interviews. This sentiment
contrasts the perspective of employees I spoke to who worked at MySpace and helped transition it
under Specific Media, who were skeptical of the groups that continued to interact on the old
MySpace platform and were wary of their style of interaction. Nonetheless, the mechanisms through

134
which Vine employees rationalized why these communities were important were similar to how
MySpace employees rationalized why content loss was trivial: that is, by drawing on themselves to
understand their users.
Vine interviewees understood the platform’s communities as important because of the ways
they personally had found online interaction to be meaningful. Becky, for instance, discussed how
she had long wanted to work for a social media platform because of how much she had enjoyed
online interaction, especially as an early Twitter user. Similarly, Caroline was an early Twitter user
and an early Twitter employee. She warmly recalled when Twitter was a lively and smaller-scale
platform, its style of interaction pervaded by a sense of good fun and good humor. She identified a
similar spirit as existing on Vine. The video platform seemed to represent social media at its best
because of the type of community interaction that took place there; you could feel like you were in a
“group of people making things together and laughing and sharing,” as she described. Because of
this feeling of community that she experienced as a user of the platform, she lobbied successfully to
make the internal switch from working on Twitter to working on Vine.
Caroline and Becky’s perspective could be understood as a nostalgic view of community on
social media platforms, in that they both understood Vine’s community through a desire for
internet-supported communities they were involved with in the past. Cultural theorist Svetlana
Boym (2001) has noted how nostalgia is a powerful sentiment, shaping interaction with the present
through a yearning for the past. In this sense, the nostalgic view of online community may have
further influenced the idea that Vine should be preserved, as it represented what otherwise seemed
to be a bygone era. Indeed, Boym (2001) describes nostalgia as a cultural feeling that occurs
especially in times of significant change, a desire to revisit a time as one can revisit a place. Both
scholars (e.g. Miltner & Gerrard, 2021) and journalists (e.g. Yalcinkaya, 2022) have observed a recent
resurgence in nostalgia for early social networks (like MySpace), which seem to represent a simpler

135
time for online social interaction, one where the internet appeared more democratic, less
standardized, and less monopolized by a few large tech companies (ibid), however untrue these
assertations may have been in practice.
Others, like Drew, had not been Vine users prior to joining the company, but embraced
Vine’s community ethos over time spent working on it. He noted how he had “come in as more of a
cold engineering person, but then I got attached to it as I spent a lot of time and effort on it.” This
attachment was particularly solidified during the course of sunsetting, as he recounted how:
People would reach out and they’d have all these heartwarming stories about how Vine
affected their life, or maybe they lost a loved one and their Vines were super important to
them. At the time [of the sunset] I was very passionate about keeping all that stuff alive.

The ways that the value of community pervaded the sunset went beyond just concern over
maintaining online communities’ content, however. Allison, a content editor let go the day of Vine’s
announced closure, noted that it was communities’ continued ability to interact that concerned her
in sunsetting. She described the range of interests of the communities on Vine, and her worry that
they had few other platforms to continue these connections on:
There’s a bunch of K-pop people. There’s people who make six second edits of music
videos. Funny things. There’s musicians. There are dogs. There are owls in Japan. There’s
this just global thing that all have their own very small subcommunities. That's probably
what made me feel the most bad when it shut down. Still now, I just think: where did all
those people go? They didn't have anywhere to go, that’s why they were here–because this
was the place that was for them.

Indeed, sunsetting a platform when it still enjoys active use brings in another element of concern for
what effect a closure has on a community. Not only are typical forms of mnemonic media stored
and organized by the platform, but so too are socio-technical connections to other users. There is
precedent for entire communities intentionally moving their locus of online interaction in response
to a site’s closure or disagreement with site administrators–an occurrence that has especially been
documented in online gaming and other niche communities (De Kosnik 2016; Pearce, 2009)–but

136
this is less common in platforms that support a multitude of user communities. While Vine
employees organized around saving content in the sunsetting process, they did not advocate for a
means to maintain diverse communities’ connections to one another: for instance, suggesting a
different platform that these users might use to continue their interaction.
It was not lost on employees that Vine served as a unique platform for groups historically
excluded from representation, and especially self-representation, in mainstream media. These
employees noted that Black content producers and Black audiences especially shaped Vine’s culture.
At the same time, the large majority of Vine employees I interviewed presented as white. As Becky
noted, “Black creators were the heart and soul of Vine. They never made money and they never got
famous, except for in their own Black community. They didn't get mainstream famous in the way
that white creators did.” She at once recognized the way that Black users shaped the culture and
constituted specific communities of cultural practice on the app, while not being supported by
platform companies or the larger entertainment ecosystem in the ways that others were. “What I
realized [while working at Vine] is Black creators were driving the culture and driving the ideas, and
that white creators were ripping those ideas off and making money and becoming famous. All of
tech is playing into systemic racism in a way that is truly disturbing.” What Becky observes is indeed
a longstanding pattern of white appropriation of Black cultural production, that has ranged from
visual arts to music to, as we see here, social media and beyond (Cashmore, 1997). Employees’
willingness to critically examine the sunsetting process, as Becky does here, was in part motivated by
the sense that the way that sunsetting was proceeding–starting with the fact that the platform was
shutting down at all–violated a responsibility to the communities that had connected there.
These perspectives contrasted the unselfconscious ways that whiteness was centered for
both Friendster and MySpace. Indeed, the demographic homogeneity of US tech cultures has been
shown to inscribe white masculinity into these cultures’ norms, in so doing creating environments

137
that are at best ambivalent and at worst hostile to those who fall outside them. Despite the
continued demographic homogeneity of participants from Vine–who, like the rest of my
participants, presented as predominantly white and male (see Appendix A)–these interviewees
showed much greater sensitivity around issues of identity, especially race, when discussing the
impact of the platform. My interviews suggest, as I noted, that employees’ investment in the idea of
user community helped them exercise greater sensitivity to social difference. At the same time, Vine
employees may have also been influenced by the larger social conversations about systemic racism
and white supremacy that were taking place as part of the social mobilization around racial injustice
that took place during the summer and fall of 2021, a time that indeed overlapped with these
interviews.
Aligning Corporate Values with Employee Values
As much as the employees I spoke to lobbied for a different vision of sunsetting than was
proposed based on valuing user communities and internet culture, the employees also explained how
they needed to rationalize this new sunsetting path by arguing that it was beneficial for Twitter in its
quest for profitability. Twitter, as it was explained to me, was “blown away” by the impact that the
closure announcement had. As Becky noted, “because they had no investment personally or in a
first-person way, they thought that nobody would care. They did not calculate the impact that it
would have on pop culture and their own product in closing Vine.”
This outpouring of public support–and the anger directed at Twitter for cancelling Vine,
coupled with the sunsetting team’s advocacy–helped change the course of Vine’s sunset. Employees
argued that they needed to keep Vines in perpetuity otherwise they would harm the larger Twitter
brand. Because Vines had been developed technically such that they played seamlessly on Twitter,
acting as an integrated video product on that platform, employees argued that removing Vines
would especially hurt Twitter. Because Twitter still maintained its user content, all old Vine posts

138
would be effectively dead, each an eyesore on the extant platform, each a reminder of Twitter’s
failure to keep Vine alive. As Becky continued, “That is what in the shutdown we lobbied so hard
for: you absolutely have to keep Vine working because you're going to damage your own product if
you don't. […] Set yourself up for success there and don't make Vines die.”
This reasoning led toward a reconsideration of Vine’s original sunsetting plan. The format
they eventually settled on was to transform all Vine content from an app to a static website, where
people could access old Vines from their account and from other users whose accounts were public.
They also transformed the existing app in to one that removed all social networking features but
maintained the editing functionalities, called “Vine camera.” Here, users could make videos that
included the quick-cut transitions that were endemic to Vine videos’ style. In 2018, when the
European Union’s General Data Protection Regulation (GDPR) went into effect, more information
was removed from the Vine archive to continue to make the site compliant (like younger European
users’ data, as well as any identifying metadata), such that what remains now are posts, captions, and
the user’s handle. Today, if you have the original link for a Vine, it will still play on a web browser or
on other platforms where it was embedded. Drew, who continues to work for Twitter, spends up to
one third of his job maintaining this static state.
Is There a ‘Right Way’?
Though all Vine employees I spoke to who were involved in sunsetting Vine expressed
similar values to one another, these sentiments were not necessarily uniform. Some employees
resented the fact that they had to stay on, to work on tearing a system apart only to still be let go
immediately after. Some engineers that were chosen to help sunset the platform, Drew noted, would
not come in to work, or spent their time holed up in the conference room, interviewing with other
companies. Two engineers quit during that time. The sunset took longer than expected. In some
ways it was a nightmare, Drew admitted.

139
Moreover, not all employees, nor users, were satisfied with Vine’s ultimate state after
sunsetting–especially because many thought Vine should have never sunset in the first place. Allison
and Paul, for instance, were not involved in sunsetting, but both thought that Vine’s sunset state did
not do justice in representing the platform. The archive was not smooth nor easily navigable, and
curated collections that Vine staff had made at the end of each year had eventually disappeared.
Allison thought that the right way to shut down the platform would have been for the sunsetting
team to work with an organization like the Internet Archive. This was for economic as well as
representational reasons:
Here's what I think–is that it should have been part of the Internet Archive. […] At least
some parts of it should have been maintained, especially the ‘Trends’ pages, of all these
things that are still today part of culture that people forgot originated there. […] It should be
part of the Internet Archive so that it's maintained by someone that are not the people who
actually paid for it.

Vine had in fact been in conversation with the Internet Archive, as Drew noted and an archivist at
the Internet Archive confirmed. However, for the non-profit organization to transfer and host Vine
data in perpetuity, they noted that they would require a million dollar in kind donation. That was too
steep for a company attempting to cut Vine because of its cost. Paul, similarly would have wanted
“some sort of curated end state,” but he lamented that things that recreated that–like the year-end
compilations–were gone. (When I spoke to him, he noted how he had recently messaged one of the
sunsetting engineers to see if he could get access to those files.) These compilations not only
represented the platform to him, but also encapsulated the cultural zeitgeist of each of those eras:
“They feel like cultural touchpoints. […] It just puts you at a time and a place.”
What the Vine employees I spoke to did agree upon was that Vine deserved a more thought-
through and user-centered sunset than it was originally planned to have. These perspectives were
ultimately captured well by Becky, who noted that what she learned at Vine was that “there are
things that cannot be replicated by tech. There are things that are important around human

140
relationships.” These attitudes embody an ethic that could be likened to stewardship–care and
responsibility for an entity over time–that shaped how Vine shut down and how it persists today, as
I explore in greater detail in the next chapter and in the Conclusion.
Conclusion: Sunsetting as a Value-Laden Act
Sunsetting is a value laden act. Employees brought personal attitudes, ideals, and worldviews
to the act of platform shutdown. As employees navigated the priorities of legal systems,
corporations, user communities, and ultimately, of themselves, they shaped decisions about how
user generated content was handled during platform shutdown.
MySpace and Vine show the range of values that are brought to this process, whether that
was valuing online community or valuing the ‘creative destruction’ of innovation. Together, these
cases show the ways that values are organized around user content when this content’s typical
economic value has been emptied. They demonstrate the ways that platform afterlives are socially
constructed––meaning that human action shaped the structure and function of these platforms’
afterlives––just as much as functioning systems are. To put it differently, how platforms will
continue to function as digital memory infrastructures–systems that underly, support, and shape
modern memory practice–is significantly socially shaped.
This chapter shows not only which values were brought to bear on this process but also how
these values were mobilized to directly shape a platform’s afterlife, what social mechanisms they
were mobilized through. The cases examined here show two main forms of mobilization. For one,
employees actively connect their own experiences and beliefs to understand who they think their
users are and what their users desire. Many studies have pointed out that the values and the
worldviews of platform designers shape the functioning of technical systems. This chapter shows
empirically the ways in which this occurs, as employees draw on themselves to inform the perceived
right way to handle something–in this case, sunsetting. The influence of employees’ values on the

141
functioning of a technological system are even more pronounced when it comes to end of life, I
assert. Because few remain at the end to shape how sunsetting occurs, individual opinions hold
greater weight. Moreover, because there are few industry rules or well-known norms associated with
the sunsetting process outside of legal parameters, these organizations had to largely rely on these
few employees and the values they brought to sunsetting. Secondly, values were further mobilized
through employees’ affective responses during sunsetting. Vine employees, upset and angry, drew on
these emotions to act as ‘rogues’ during the sunsetting process. MySpace employees, in turn,
responded to a feeling of excitement over the new platform they were building, further entrenching
the old platform’s content as burdensome and dull.
At the same time, sunsetting was influenced by the values of the organizational culture at
large, as well as the types of values that were fostered around each platform over time. MySpace’s
new guard, who I focused on in this section, had thought highly of MySpace’s cultural cache in its
early years, the artists it supported and the kinds of connections it had made. But they largely viewed
any remaining users, and user content that remained, as outside of this cultural vision. Instead, they
aimed for the creative destruction attendant to ideologies of innovation. Vine, in contrast, had in
general thought highly of its user community and their cultural production over the course of Vine’s
tenure, though they resented their positioning within the larger Twitter organization. This
organizational culture logically underlies the dynamic that existed at end-of-life, wherein content
continued to have intangible cultural value to employees in the sunsetting process even if this was
not the prevailing view for Twitter executives.
As was noted at the beginning of this chapter, Hawkins and Muecke (2003, p. x) have
ventured that social values are subject to reorganization when waste is subject to non-circular modes
of valuation, that is, when objects escape straightforward regimes of production and consumption

142
and disposal.
19
In the case of sunsetting, this would mean that employee values and organizational
values may shift in the sunsetting process if user-generated content and user data function outside of
these regimes. What was seen through the cases explored here, however, was not so much the
complete reorganization of values as it was a dynamic in which previously aligned forms of valuation
(ascribing worth) and social values (what individuals deemed important in life) became misaligned.
To take Vine as an example: prior to the decision to sunset the platform, the economic
worth of user-generated content aligned with social values around Internet-based community and
Internet-based culture. Maintaining a platform for this community and for their cultural production
was supported by corporate forms of valuation that ascribed economic worth to user content.
Hosting user content kept users engaged and supported a corporate bottom line because they could
advertise to these users (or at least had the promise of turning a profit through advertising to them).
However, when sunsetting occurred, values around community and cultural production fell out of
alignment with these corporates processes of valuation, which, without the promise of advertising
profit, resulted in desire for its disposal. Values were not always reorganized in these cases, then.
Instead, they were subject to reassertion, as Vine employees brought these values to the fore to steer
decision making in sunsetting. They could also be subject to realignment, as groups attempted to show
how these social values matched up with ongoing dominant processes of valuation.
In MySpace’s case, corporate forms of valuation and employee values were largely aligned–at
least in the cases of those I spoke with. However, value and valuation misalignment may very well be
at the root of the discomfort, ambiguity, and irony that were seen in these conversations. MySpace
employees’ valuation–the designation that the bulk of user data had little economic nor cultural
worth, or that particular communities were more desirable than other communities–were misaligned

19
It is worth asking whether any object subscribes fully to these ‘straightforward regimes,’ and how objects
relate to the organization of social values if this is the case.

143
with values popularly associated with platforms and web 2.0. Long after the alignment in values and
valuation that marks the auspicious, hockey-stick growth days of these platforms, sunsetting
provides a glimpse of this relationship when things start to fall apart. In doing so, sunsetting
highlights how values operate––and importantly, that values operate––in platform cultures beyond
these early promising stages.

144
Chapter 4
The Construction of Destruction: Material Considerations in Sunsetting
My interviewee had made it abundantly clear: constructing something that will need to be
deconstructed is antithetical to the practices and perspectives of most forms of engineering.
Architect at Friendster Peter had vividly contrasted engineering, whether computational or
structural, to the Tibetan Buddhist practice of building sand mandalas. Intricate, colorful, geometric
and painstakingly crafted by monks using tools known as chak-pur, they are built only to be
ritualistically destroyed after completion, the sand traditionally thrown in a river (Figure 6). The
process symbolizes the transitory nature of material life. “The whole point of when Buddhist monks
do sand mandalas is that they are going to throw them in the river,” he explained. “That’s just not
the general ethos of engineers.” Instead, the engineering ethos he described is one aligned with a
focus on innovation, meaning the designing of the novel, as well as one that is based around tenets
of stability, permanence, and reliability. Just how antithetical is the act of dismantling to the ethos of
engineering? Peter suggested that it may require hallucinogenic drugs for engineers to shift their
perspective so fundamentally during the process of design as to understand that one day their
systems would need to be deconstructed and destroyed. This chapter focuses on times when
engineers and other employees from social media platforms had no choice but to confront the
process of destruction and dismantling.

145

Figure 6. The process of creating and destroying a sand mandala.
Screenshots from Wheel of Time (Herzog, 2003).

While the previous two chapters have traced shifts in valuation as a platform goes into
decline, and the personal values that employees bring to the decision of what to do with user-
generated content, here I focus on the socio-material processes of sunsetting, or what could be
called the construction of destruction. That is, employees grappled not only with personal, professional,
and organizational values in designing platform sunsetting, but also had to grapple with technical
considerations–from the original structural configuration of the site to the variable difficulty in
maintaining physical storage infrastructure to the challenges in dismantling interrelated software
components. As this chapter shows, the priorities, logics, and goals expressed by these actors in
combination with material considerations ultimately shape the magnitude of commercial digital loss
and the afterlife of digital memory infrastructure.
The Construction of Destruction: Between Materiality and Constructivism
Framing this chapter as focusing on materiality during sunsetting while at the same time
titling this inquiry the “construction” of destruction inevitably leads toward a question of approach:
between materialist and social constructivist approaches to technological systems, as well as

146
materialist and symbolic approaches to media technologies in the context of communication studies
(Gillespie, Boczkowski & Foot, 2014). While materialist approaches to technology tend to focus on
artifacts’ physical nature as an important component of their interaction with social life, social
constructivist approaches to technology stemming from sociology (Berger & Luckman, 1966/1991)
and developed in STS (e.g. Kline & Pinch, 1996) argue strongly against forms of technological
determinism, instead examining how social vectors shape the creation, use, and societal roles of
technology. Of course, approaches that draw on both are widely recognized, from actor network
theory’s arrangement of human and non-human relations (Callon, 1984) to the widely used but non-
trademarked notion of ‘mutual shaping.’ Scholars have also discussed that there is heterogeneity
even in what appears to be the poles of the social/technical spectrum, with concepts such as “hard”
and “soft” determinism (Smith & Marx, 1996), or “radical” and “mild” social constructivism
(Sismondo, 1993).
One must stand somewhere. In line with dual approaches, this project understands
sunsetting as a process comprised of both social forces and material forces. In emphasizing
materiality in this chapter, I focus on “the physical character and existence of objects and artifacts
that makes them useful and usable for certain purposes under particular conditions,” as Leah
Lievrouw has characterized (2014, p. 26). Importantly, though the previous two chapters focused on
‘social’ vectors like valuation and personal values, they were never divorced from material and
technical considerations. For instance, the notion of Southeast Asian Friendster users being
“resource-intensive” population was inseparable from specific technical configurations related to
scaling the platform’s computational power; the ‘social network contract,’ or that commonsense
notion of content maintenance discussed through MySpace, was linked to a platform’s storage
configurations. In turn, bundling material concerns in this chapter is not intended to isolate the
material from the social but rather to analytically emphasize materiality at a moment in which

147
material concerns became most prominent in interviews with platform employees. As will be seen
here, as platform organizations decided what to do with user data, employees both confronted and
discussed the materiality of their respective platforms, shaping the actions they took vis a vis content
preservation or deletion.
By discussing destruction in terms of construction I not only mean to signal an alignment
between the material and the social, but rather play with the practical image of construction: people at
work, putting pieces of material together according to a design, creating some larger whole, or, as
the Oxford English Dictionary defines it, “the action of framing, devising, or forming, by the
putting together of parts; erection, building” (construction, n.d.). As was shown in chapter 2, as
platforms went into decline, organizations moved from an understanding of their business as an
interrelated and interworking system to one that was made of components with variable extractable
value. By ‘piecing apart’ (instead of piecing together) their platform, platform sunsetting comes to
resemble construction, in which readymade parts stand alone, ready to be configured. The image of
construction also puts an emphasis on the material practices of employees, and the ways that these
material practices move toward some stabilized whole. Of course, when understanding the
construction of destruction, the stabilized whole in this case is somewhat paradoxically an absence,
that is, the loss of user-generated content. In focusing on construction, this chapter argues that
absence, deletion, ruin, are not outside the mutually shaping forces of materiality and sociality, but
rather part of a process of material and social structuring. In the next section, I expand on the
construction of destruction by focusing on materiality in software specifically, connecting insights
from software studies to sunsetting.
Dismantling Software Systems
In earlier chapters, I noted that there was limited inquiry into the dismantling of digital
knowledge infrastructures despite sustained inquiry into the maintenance and breakdown of systems

148
more commonly identified as infrastructure, like roads and bridges and power grids. While literature
on this subject would be instructive in understanding the sunsetting of social media platforms, for
understanding materiality in this context research from software studies on obsolescence is
enlightening. When looking to the material stuff of platforms, software is indeed critical: code is a
major part of what constitutes a social media platform. At the same time, software systems are
“more than the text of code,” as R. Stuart Geiger has noted (p. 347), in that they are inseparable
from material artifacts, practices, and people. Platforms likewise are partially constituted by software
but only come into being through that software’s relationship to hardware, physical components
from the circuit board to the server farms. The centrality of software to platforms, however, allows
generative lines to be drawn between observations from software studies toward sunsetting.
Wendy Hui Kyong Chun (2011) has described software’s paradoxical ontological status as
that of the “enduring ephemeral,” meaning that software which does not change, which is not
upgraded and updated, does not endure. This contrasts, say, a traditional infrastructure like a bridge
or a traditional medium like paper, contexts where if an item remains unchanged then it endures.
Software, on the other hand, becomes obsolescent in part because of its entanglements with
interrelated systems, whether a scripting language or a computer’s operating system. Update these
interrelated systems and the software in question will not endure in perpetuity. Chun notes that
software therefore presents an ironic relation to common perceptions of digital media, which have
been heralded as saving other forms of media from destruction in their ability to converge physical,
‘bulky’ media into digital storage (for instance, that a printed magazine will be able to persist because
of its archives’ digitization instead of the need to perpetually store original copies). What Chun’s
“enduring ephemeral” presents is not so much an unprecedented ontological status for technology
based on these modes of decay, but rather the complex and interdependent material relations that
software is enmeshed in.

149
Christopher Kelty and Seth Erickson (2015) frame Chun’s “enduring ephemeral” as an
analysis of software durability, and similarly differentiate the required maintenance of software from
the maintenance of more traditional infrastructures like roads and bridges. Importantly, they
highlight the ways that keeping software running relies on webs of dependency:
Maintenance of software, as software professionals often recognize, is not quite the same as
maintaining a bridge or freeway: it is not about wear and tear or the failure of particular bits
of software. Rather it is about keeping software in synch with changes and dependencies
made in other software and systems.

By highlighting software’s fragile forms of entrenchment and interdependency, Kelty and Erickson
foreground software as a maintenance-intensive artifact, an entity requiring vigilance and awareness.
If not, software becomes subject to what developers call “code rot,” the degradation of code by
virtue of its changing environment, not by any physical degradation to the code itself. Observing
these constantly shifting dependencies, Marissa Leavitt Cohn has framed software as a “timely
object.” She notes that a sensibility toward temporality should guide analyses of software, arguing
that (2019, p. 424):
… the significance of software’s materiality for the politics of computational work lies not
only in its present performances—however hard these may be to trace—but also in its forms
of duration, the entangled lifetimes of careers, professional identities, and programming
languages and paradigms, all of which come and go. This of course still requires an STS
understanding of code as a relational assemblage but also emphasizes that these
entanglements have duration […].

The material approach to platform sunsetting explored here attempts to build on observations of
software’s complex interdependencies and particular modes of obsolescence and degradation, but
especially aims to highlight these interdependencies with an eye toward duration, toward shifting
material entanglements over time.
Given this complex set of relations that literature on software maintenance and breakdown
presents, I came to interviews with platform employees curious whether the act of sunsetting was in

150
fact technically simple. Did it resemble the ritual of building and destroying a sand mandala, with the
construction of the system layered and complex but the act of destruction seemingly sudden, swift
and easy? Does it happen in the way that it appears to the outside viewer, with data suddenly gone as
soon as the plug is pulled on servers? Or does the process resemble something more like
construction, with systems coming into a state of destruction, piece by piece? On the one hand,
previous chapters have shown that sunsetting and commercial digital loss do not happen all at once:
processes of valuation and personal values shape platforms and their sunsetting over the long durée.
Focusing on the material aspects of sunsetting, on the acts of dismantling and content removal, as
I’ll explore here, however, paints a picture of sunsetting that highlights the relative swiftness of
content deletion if the decision to remove content is made.
MySpace: Technical Debt to Total Destruction
This chapter will conclude MySpace’s story. To do so, it addresses the platform’s massive
content loss in 2019, perhaps the point at which the sun seemed to ultimately dips below the
horizon during sunsetting. But it also reflects on material considerations from the early construction
of the platform under eUniverse and Intermix, as well as how this technical substructure ultimately
shaped the deletion of content in 2013 under Specific Media. Indeed, to understand MySpace’s
content loss in both 2013 and 2019, one must return to the very beginning of MySpace and its
origins in the “churn and burn” culture of eUniverse.
‘Bubblegum and Tape’
Even if MySpace was not built to be destroyed, it was also not built with longevity in mind.
That “churn and burn” ethos that defined eUniverse, after all, was a quick and scrappy and
speculative form of development built around figuring out what could gain momentum and turn a
profit quickly. Aaron, the board member of eUniverse, described his perception of the flimsiness of
MySpace’s technical underlayer under their ‘churn and burn’ focus. He noted that MySpace’s

151
founders were dismissive of what others saw as important technical foundations for scaling, a
contrast that was especially stark in 2005 and 2006 when compared with upstart Facebook. “‘Guys,
but they are building scalable technology,’” he said, referring to what he thought of saying to
MySpace leadership about Facebook:
“Over time you’re going to have to think about [that too].” MySpace wasn’t built for that. I
know this because I'm Silicon Valley through and through, I grew up around here. MySpace
was built with bubblegum and tape and not really with scalable technology infrastructure.
[…] The engineering team was lean, and not very good. They were more about the branding
elements and like, “We're going to be this great site for musicians and Hollywood stars.”

The discrepancy between the “cultural” focus of MySpace and what Aaron saw as robust “technical”
practices shows just one way in which material considerations and social vectors are deeply
intertwined in the building, operating, and sunsetting of a platform. Employees like Nicole likewise
discussed how, while Silicon Valley techno-cultures tended to divide large technology companies as
either focused on “product culture” or “engineering culture,” MySpace was anomalous in that they
focused on “creative culture.”
Aaron’s description of MySpace as built on “bubblegum and tape” was not the only time I
heard that phrase. Peter, the architect at Friendster who contrasted engineering with the process of
building sand mandalas, had also used it when describing the only time that he believed engineers
designed with temporary construction in mind. He noted that the one exception to the dominant
engineering ethos was developers who specialize in creating sites and services for political
campaigns, who approach these systems with an eye toward their eventual dismantling when the
campaign ends. Software engineering in these contexts is, as he noted, “pretty wild because you’ll
make decisions that look a lot more like–an analog might be bubblegum and duct tape. These are
things that you would never do if you if you were [working] anywhere other than a political
campaign.” MySpace’s early architecture was therefore defined by a material structure thought of as
one held together by quick fixes, one not built to last.

152
These conditions helped create what developers both inside and outside of MySpace call
“technical debt,” a concept commonly defined as the “accumulation of all the technology work a
company needs to do in the future” (Dalal et al, 2020). More concretely, technical debt is analogous
to financial debt, in that it is typically caused by making technical shortcuts or good-enough fixes to
finish a working product while still leaving a list of to-dos that, like financial debt, that will one day
need to be ‘paid back.’ Moreover, technical debt is often linked by developers (in both my interviews
and in grey literature) to the use of “legacy technology,” or systems that are deemed to be
obsolescent or outdated and in need of updates or replacements. As I’ll discuss in more detail soon,
technical debt is an amorphous concept, but MySpace developers recognized specific modes of it as
a factor in MySpace’s data deletion in 2013 under Specific Media.
What’s a Platform Made Of?
Before diving into the scope of MySpace’s technical debt, it is worth pausing to ask: What
constituted MySpace’s “bubblegum and tape?” What material substructures were MySpace–and
other platforms I examine here–composed of?
The sites I describe in this project are made of numerous technical layers of hardware and
software.
20
As in Kelty and Erickson’s (2015) description of software interdependency, design
choices made at different layers in this architecture interact with and influence one another. More
specifically, these platforms are composed of two main “stacks,” named as such because of this
layered architecture. The first stack relates to the internet and computer networking and is
composed of software layers in both clients and servers, which ultimately all aid in building the
infrastructure for communication between hosts (computers) defined by internet protocols like
TCP/IP.

20
Information for this section comes primarily from an interview with Jeremy, a technical lead at MySpace.
Thanks to Alex Konrad for help in technical translation.

153
The second stack is the software stack, composed of different pieces of software each with
their own purpose, which, interlocked and interrelated, perform the functions of the platform. This
stack includes the database layer, which allow for the storage and management of data. While
databases are storage infrastructure, they do not store what one usually thinks of as user-generated
content. Instead, databases store symbolic data like text or numbers, as opposed to multimedia data.
The software stack also includes servers, typically housed at off-site locales but sometimes at a
company’s office, as they were with earlier sites like GeoCities and Friendster. Servers work with a
diverse array of platform activities, handling user traffic, running the code for the platform
application, and marshalling the file system where user-generated content resides. Files may in turn
be stored on those servers’ hard disks or on storage architecture like a SAN; today it is more
common to use cloud storage, which is cheaper and more accessible as a service. Finally, a platform
typically interfaces with third-party CDNs, or content delivery networks. CDNs were first proposed
in the 1990s as a solution to network congestion caused by media-rich and data-heavy content like
graphics and videos (Dilley et al, 2002). In CDNs, geographically distributed servers cache this rich
user content, allowing it to be served up more quickly to nearby users who want to access it.
MySpace was originally built using ColdFusion, a software development platform that
allowed the company to come to market rapidly (Wunker, 2011). ColdFusion in turn interfaced with
a large Microsoft SQL database layer, which ran on a combination of servers in both Microsoft and
Linux environments. Employees noted that the combination of ColdFusion and their proprietary
database layer made the platform both expensive to scale and unwieldy over time, especially as
ColdFusion was “a very rudimentary programming language that was never intended to scale to the
size MySpace did,” as Brendan noted.
However, the technical debt that MySpace accrued was not necessarily because developers
made early mistakes that contradicted industry best practices for scaling a social network. Instead,

154
employees discussed that there were few known precedents during the era of MySpace’s debut for
scaling a site so quickly, for handling the crush of user traffic and user-generated content. “We called
it ‘changing the tire while the car was driving down the highway,’” Louis said of MySpace’s rapid
growth during 2005 and 2006. “There's nothing we could do. We have to keep moving forward. We
have to just figure out what the issues are, just fix it as fast as possible.” Like Aaron who described
the platform as built from bubblegum and tape, a developer who joined the platform under Specific
noted that MySpace’s architecture “[seemed] crazy.” However, when looking at the specific technical
contexts and era in which it was developed, you started to see the reasoning, eventually seeing how
“oh, yeah, that makes sense.” Moreover, platforms like MySpace would invest in massive amounts
of a particular hardware to scale to the size they needed. However, these parts could become quickly
obsolescent, with cheaper and more efficient resources coming on the market. A platform’s
hardware resources would then be both outdated and un-resalable, making fundamental technology
change a poor financial decision for the platform.
21

MySpace eventually moved from ColdFusion (later porting to ASP.NET, an open-source
web framework developed by Microsoft [Kraus, 2006]). Nonetheless, the implementation of these
early technical arrangements would pose problems as MySpace attempted to transform themselves
after their decline.
Technical Debt as Installed Base
The idea of “technical debt” pervaded my discussions with employees who worked on the
technical side at MySpace. Early employees from MySpace’s debut, from its peak, from the era of its

21
This dynamic was especially apparent for Friendster. As international development manager Steve
described for that platform: “We had committed ourselves to a server architecture […] that was four times as
slow and cost four times as much as what Facebook and other later social networks were able to [have]. They
actually learned from the multimillion dollar mistakes that we made. And that was partially a function of that
the technology improved during those three years. By the time the newer, cheaper, faster service came out, we
were already built out, fully built on legacy architecture, and it would have been catastrophic to transfer
everything from one server cluster to another.”

155
decline, and finally those who joined during the platform’s ownership by Specific, each discussed it.
MySpace’s accumulated technical problems made the platform unwieldy to work with and especially
resistant to change. Krish, for instance, had joined MySpace when his company was acquired by
MySpace and News Corp in 2010. The MySpace organization had wanted to absorb features from
the product he and his team developed. Krish recalled how the team that was acquired worked
separately from MySpace’s existing technical base because of how difficult it would be to integrate
the new products they were building into MySpace’s “legacy” technical architecture:
The stack that we had built prior to MySpace was at the bleeding edge of technology.
MySpace’s technology stack was the exact opposite of that, like pretty much the oldest
technology you can have and still be viable at that kind of scale. […] Everything that we built
for MySpace, we built completely separately from the rest of their technology, because it just
didn’t make sense for us to attach these modern tools to this legacy stack.

MySpace’s technical debt that resulted from its “legacy technology” could be thought of as being the
effect of an installed base (Star & Bowker, 1999; Star 1999). Stemming from studies of knowledge
infrastructure, an installed base refers to an infrastructure’s relationship to existing sociotechnical
structures. Infrastructure, rather than developing anew, instead “wrestles with the inertia of the
installed base and inherits strengths and limitations from that base” (Star, 1999, p. 382). When those
building an infrastructure do not understand the constraints of that base, it can hinder new
development of the infrastructure in question (ibid). MySpace could very well be seen as a case of
this misrecognition, based on employees’ characterization of MySpace leadership as not recognizing
the importance of building robust technical architecture. At the same time, it shows the ways in
which platforms grapple with the durability of design choices made earlier on. As Jeremy, the VP of
technical operations for MySpace, discussed, the original architecture was stable enough but rigid,
and it set up a tradeoff, especially during the transition to the new MySpace, of maintaining user data
or re-architecting the platform.

156
There was that rigidity that was stuck in there. Yeah, you have technical debt. How much do
you want to spend to relieve yourself of that technical debt? It was a lot of data when you
have all of those custom profiles to maintain. A lot of data. To make any changes to [the site
architecture] and maintain people’s data, it was very, very, difficult to overcome.

The MySpace organization was hamstrung by these technical foundations both in MySpace’s hockey
stick years and in its years of decline. In the early years, employees were already overwhelmed trying
to patch the existing architecture to keep the site online, let alone having the time to re-architect the
platform. During its decline, there were fewer financial resources that could be allocated for a
complete rebuild. That is, employees from later eras of MySpace had to contend with the installed
base that earlier iterations of MySpace employees had put in place.
Moreover, some employees phrased the accumulated data as one form of technical debt.
When I asked Troy, a database administrator for Specific who focused on MySpace after the
acquisition, what he meant when he used the term “technical debt,” he described it as such as “like
your garage. You build all this stuff, and you keep filling it up. Over time, it just gets so cluttered and
yet, what are you going to do with it? Move it or manage it or delete it or throw it away?” Troy was
speaking in the context of accumulated corporate data, whether logs of accounts receivable and
accounts payable for brands that no longer existed, or 20,000 employee email distribution lists when
they only had 1,000 employees. He noted how:
People are afraid to delete one, because if we delete one, and then there's somebody that is
trying to send an invoice there, then it gets bounced, and that whole process breaks down.
So you just keep it, and you end up just keeping it, just keeping it, just keeping it and all of a
sudden, you're paying for all this infrastructure because you're afraid to shut it down.

The flexibility of the term “technical debt” illustrated in Troy’s description, its ability to encompass
technical to-dos as well as the burden of accumulated data, suggests that old user-generated content
would be subject to the same logic. That is, it could easily be framed as a costly encumbrance that
accumulated over time and that needed to be resolved through deletion.

157
Employees working to build the new MySpace under Specific Media made choices around
sunsetting in part based on the difficulties of working with MySpace’s installed base. Just as Krish
described it being easier to build products that were separate from the existing architecture, that is
how Specific Media ultimately approached the new MySpace. They would not wrestle with the
interdependencies and intricacies of what by then was a decade old platform. Rather, they would
begin anew as best they could, and attempt to shake off the platform’s installed base. What they did
was built a separate site. Eventually this site would be connected to MySpace’s URL. At first, writer
Julian remembered, they had the new MySpace at a URL beginning with the prefix ‘new,’ as in
‘new.myspace.com.’ (This was likely while the site was in beta, though the exact timing wasn’t clear
from interviews.) Eventually the new prefix was removed, and the new MySpace was accessed
through the ‘regular’ URL.
One of the effects of moving to a new site and away from the masses of technical debt that
the old MySpace architecture represented was that the old user-generated content was not
automatically associated with the new architecture. Because of this, one could say that Specific
Media did not so much choose particular segments of user-generated content to delete from the old
MySpace as they did select particular segments of user-generated content to bring over to the new
design. Two things made the cut in the transfer: photos and lists of friends. As an update posted
during the transition under the title “Where’s My Old Stuff?” read (‘Where’s My Old Stuff,’ 2013;
see also Figure 7):
22

We completely rebuilt MySpace and decided to move over some of your content from the
old MySpace. What's available? 1. Photos 2. Friends […] If you don't see any photos, that

22
It is surprising (and, to this analyst, seemed incredible) that MySpace’s FAQ pages from their transition are
still available given the massive loss MySpace sustained in 2019. On the other hand, these pages may still exist
because they are lightweight, contain primarily text and were likely hosted in a different configuration than
the lost, media-rich user-generated content. Indeed, Anne Helmond and Fernando Van der Vlist (2019) have
noted that grey literature from platforms like these pages are important and relatively robust resources for
those engaging in internet historiography, a claim that is backed up by the availability of these MySpace pages.

158
means your old account was not synced to your new MySpace. Try doing a search to see if
you can locate your old MySpace account. Unfortunately, if you cannot locate your old
profile we will be unable to assist with retrieval since the old MySpace was never transferred
to the new MySpace. […] If you are looking for something that is not on this list, that
means the content is no longer available and cannot be retrieved.

159

Figure 7. MySpace's FAQ for lost content under the transition to the new MySpace (2022/2013).

160
Digital: Very Expensive, Very Material
“Software is not immaterial—this much is clear to anyone who studies it,” Kelty and
Erickson (2015, p. 5) write in their text on software interdependency. The same can be said for those
who work on platforms. What Kelty and Erickson are referring to is the widespread cultural idea
that the digital is “immaterial,” an idea that tracks with fantasies of internet-enabled disembodiment
and longstanding ideas that information is free-floating, untethered to the material world, vapory
(Chun, 2011). In general, this perception tracks with the seeming invisibility of software’s material
components. However, as I’ve argued earlier in this project and as numerous analysts have noted,
visibility in general and especially the visibility of technical infrastructure is always culturally shaped
and dependent on positionality (Sandvig, 2013). The visibility of the stack became evident in my
conversations when employees displayed a concentrated emphasis on these elements when
recounting their time spent working on the platform.
In the case of MySpace, the materiality of the “stack” was typically hyper-visible to
employees who worked in a technical capacity. They were especially visible to these employees, and
sometimes had more widespread company visibility, when elements of the stack were broken. The
most common context for this was when the server capacity of the platform could not handle the
volume of users that were attempting to access the site, resulting in servers “falling over,” as
employees called it, meaning that they became overloaded and crashed. Across all companies I
examined, these were issues that occurred especially for Friendster and MySpace, but less so for
Vine and other platforms like 43Things, Couple, and Foodily, all companies founded later that used
cloud-based infrastructure rather than on-site servers to scale. Indeed, in the case of these cloud-
based companies, the material infrastructure of the platform was much less a point of discussion,
even for technical employees.

161
I found the adage that ‘software is not immaterial to those who study it’ or in this case ‘those
who work with it’ to be especially true when employees were working on a platform with declining
or ambiguous economic value, the case of MySpace from 2008 onward. In these instances, the
recognition of the materiality of the platform went beyond being a focus of the technical staff, those
who were ‘close to the metal’ (Brunton & Coleman, 2014). Those from the product and creative
content teams were acutely aware of it as well. Julian, the writer for the new MySpace, explained his
perspective on the platform’s infrastructure during that transition:
All this content that gets out there, it’s got to go somewhere. Yeah, it’s digital. But it has to
sit on a server. And that server has to be somewhere. And that server has to get powered by
electricity. That costs money that gets paid by: who? And what happens when something
stops making money, enough for it to pay for those servers? I think those were some of the
questions that were happening.

Julian’s description of the questions that were being asked as Specific designed the transition from
the old MySpace to the new MySpace gives a sense of how the materiality of a platform is intimately
tied to processes of economic valuation. That is, one of the ways that the materiality of a platform
becomes visible across an organization is through its expense–its monetary expense, yes, but also the
fact that it needs human resources to support it, as will be discussed in the case of Vine. (Of course,
these employees also represent economic cost: as a MySpace employee estimated, IT costs make up
15% of the overall budget of a platform, but salaries are a company’s most significant expenditure).
In the end, this close association between the material and economic requirements of a platform
shaped sunsetting, as storing content on a server that required a physical location and electricity and
maintenance also needed to be supported by financial resources.
Commercial Digital Loss in 2013
These factors ultimately led to the removal of user content as MySpace made their transition.
What did this look like on a material and technical level? No person I spoke to was privy to, or
perhaps cared to share, what it looked like to actively delete data from the old MySpace in 2013. But

162
interviews with employees from other platforms who had engaged in the technical deletion of
content were enlightening. While each platform necessarily deals with the design of the technical
architecture in sunsetting and removing content, there are some things that appeared across
numerous instances of sunsetting. These practices were likely similar to how MySpace proceeded in
that era.
Many of the platform employees that I spoke to say the first step was always notifying the
users that the service would be shut off. Typically, people may download their content if the
platform has given them a path to do so, but eventually traffic dwindles. As traffic dwindles, the
employees move the platform onto progressively cheaper hardware and consolidate storage on
cheaper disks. This involves a lot of copying between servers, until the point at which everything is
shut off.
Once the shutdown date is reached, rather than just pulling the plug on the infrastructure
immediately, platforms typically disconnect or redirect the URL from the page’s hosting instance. In
the case of MySpace, instead of the MySpace.com leading to the ‘old’ MySpace, the ‘old’ MySpace
would still exist but would have no pathway for user traffic to reach it, while the ‘new’ MySpace
would be linked to the original MySpace.com URL. As Gibby, the founder of music-oriented
platform Makeoutclub explained, once the URL is redirected the shutoff doesn’t happen all at once
for everyone around the world. Rather, as he noted, it depended on IP address and physical distance
from the site’s cached data stored on its servers. “It takes a while for it to matriculate across the
internet, depending on what your IP is, depending on your distance from the data and the registrar.”
As he shut down the site on New Year’s Eve of 2016, he described it as percolating across the web
like a wave: “some people were like, ‘Ah! It’s gone!’ and then some people were like, ‘I’m still using
it. I’m in Thailand,’ or whatever.”

163
Once traffic has stopped completely, employees are able to take the infrastructure apart.
Servers are turned off, and repurposed–as they likely were at MySpace for the new site; sold through
a liquidation company for another company to use; or if sufficiently obsolescent, sold for salvage.
The work is still not done. To reuse, repurpose, or salvage disks, they need to be wiped of all user
data. Not doing so proposes major security risks as well. “Wiping disks,” a former Friendster
engineer recounted, “takes a long period of time, and it could break in the process, so you [may]
have to start the process over again to make sure that disc is wiped.” Files can technically be deleted
from a disk by removing the connection between a given file and a given file name, essentially
“marking” the file as deleted without destroying it. This practice is convention because of its relative
speed and efficiency. In the case of sunsetting, the disks need to be properly wiped through a
process of reformatting, or ‘writing over’ the existing data. One engineer at Friendster said the most
secure thing that can be done is to then drill through the disks after this, making them non-
functional, before sending them off on a guarded truck to a facility to be melted. (No one described
what happens after this point, highlighting an even less visible stage of dismantling. One possible
and likely path for these components is to be sent to an e-waste dump, often posing environmental
and health risks to local communities [McElvaney, 2014; Slade, 2007]).
The greatest difference between platforms’ technical process of sunsetting was based on
how those platforms had originally scaled, whether via infrastructure on premises, as MySpace did,
or those that used cloud-based services like Amazon’s AWS, Google Cloud Storage or Microsoft
Azure. In the case of a cloud-based platform, an engineer would first issue individual delete
commands to each of the files stored on cloud infrastructure. Then, through the cloud service client
panel, you decommission your cloud servers, and payment for these services is then removed from
your bill.

164
While one of the tasks under Specific was re-architecting the MySpace site, moving to non-
legacy hardware that was both cheaper than its original architecture and more efficient, the cost of
maintaining the old MySpace’s user data, the challenges in transferring this content, along with lack
of will to maintain that content, resulted in the eventual deletion of most user content in a process
like those described here in 2013. Crucially, while engineers emphasized that the process of deletion
was not quick, especially in the case of on-premises architecture, it is strikingly fast when compared
to the time it took to build and maintain that architecture by the platform, or the overall time that
users spent uploading and storing media on the site. Indeed, it was ten years after MySpace launched
that this data deletion happened, a technical process of commercial digital loss that likely took
weeks.
What Happened in 2019?
While employees hoped that removing this content would spell an easier technical and
cultural transition–not having the baggage of the old site to contend with–in the end, the moonshot
project that MySpace employees envisioned under Specific Media never reached the heights they had
hoped for. The content, and the redesign, and the launch, with an expensive but slick ad played
during a major US basketball game, at first created some momentum. But employees and the
company itself rapidly realized that their rise was more modest than they had hoped. Watching the
numbers, as Julian explained:
We had a quick moment of rocket ship and then it was like, ‘Oh shit, we got to keep doing
stuff. We got to keep finding ways to goose these numbers.’ […] I vaguely remember they
started to level off. It was still rising, but it wasn't this continuous exponential hockey stick
that kept going.

He left MySpace soon after. Others were either laid off or quit. Gavin was one of the last employees
working on MySpace. He described the last nine months of his time there as “we kind of bumped
along for a little while, then they literally fired everybody except for me […]. There were massive

165
layoffs where they laid off 75, 80, 90% of the creative team.” Specific then shut down MySpace’s
office in Beverly Hills, and any remaining employees–Gavin estimated there were around 20 left–
relocated to the Specific office in nearby Los Angeles’ Marina Del Rey. Gavin eventually left. The
internal dynamics of what happened after that are, to this analyst, reduced to speculations of
employees, popular press treatments, and chatter on Twitter.
The site stayed online, though. MySpace.com is still an accessible URL today. Employees
conjectured that it must have still drawn traffic and therefore some user data and potentially ad
revenue, so Specific kept it online. Troy, originally employed at Specific but who had joined
MySpace’s ranks during the acquisition, thought the site must have been kept online because
infrastructure developed for MySpace was used by the company’s other holdings. Specific Media
eventually turned into Viant Technology, a company that sold user data. Viant was then purchased
by Time Inc., the publishers of TIME magazine, in 2016, which was then sold to Meredith
Corporation, a broadcasting conglomerate, in 2018. In 2019, Meredith sold the equity stake in Viant
back to an existing Viant Technology LLC run by the Specific Media founders, Tim and Chris
Vanderhook (‘Meredith Corporation Sells Equity Stake,’ 2019). Some employees thought MySpace
was simply forgotten as it changed hands from holding company to holding company.
By 2019, MySpace, if indeed left untouched, would have had major technical vulnerabilities.
Those vulnerabilities may have caused the site’s massive data loss in 2019, wherein all user profiles
uploaded between 2003 and 2015, as well as all music tracks, were deleted in what the company
announced as a “botched server migration” (Porter, 2019). Even with the company’s attempt to
resolve technical debt under Specific, employees still described the platform as technically messy and
volatile. If MySpace had been maintained, employees argued, it would have only been to move it to
cheaper hardware with fewer backups, therefore reducing its price as much as possible while likely
exposing it to technical vulnerabilities.

166
Nonetheless, employees who had not known about MySpace’s later data loss were indeed
surprised, stunned even, when I told them what happened in 2019. Those that had known before
our interview expressed that they were shocked when they first heard the news. The shock was most
palpable in technical employees. Louis recalled how he had discussed it with another former
employee in 2019. His first reaction was disbelief: “I don't think that's really true. I don't know what
‘lost data’ means.” Then they “just laughed about it. We were like, ‘How the hell does that happen?’”
I indeed wanted to know how the hell that could happen. It was a question I asked all
technical employees I spoke to across all platforms, if not data loss in the case of MySpace then
what data deletion and data loss looked like in general, technically speaking. Louis, still in disbelief,
surmised that MySpace’s 2019 loss could have been caused by the move to cheaper hardware.
I don't know what it means that something is ‘completely gone.’ I don't know what that
means, because all that data is distributed among several different databases, several hundred
databases. That means that they lost several hundred different files. It was designed for that
not to ever happen. I mean, obviously when I left 'til now, they did a bunch of cost-cutting
measures to make the maintenance of the site extremely cheap. Maybe that's what happened
is they just made a mistake. Going from something expensive to something cheap, maybe
they lost a bunch of data.

Other technical employees sketched a similar scenario, though their answers were always marked by
the same disbelief, non-comprehension at how something like that could happen.
Krish suggested another theory, one that had also been voiced by a technology journalist
Andy Baio who had originally broke the story of the data loss: they thought it was intentional. Baio,
in a 2019 tweet, was skeptical that the loss was an accident. “Flagrant incompetence may be bad PR,
but it still sounds better than ‘we can’t be bothered with the effort and cost of migrating and hosting
50 million old MP3s.’” Krish noted how that “can’t happen to a real company.” He continued,
noting that having abandoned their social networking and music service prospects, MySpace was
essentially “a company that does stuff with data. Whenever you store data, you have a fresh tinder of

167
lawsuits.” Lawsuits arose because of inevitable security breaches, data and identity theft, or other
data-related incidents. He wondered if “they were trying to remove all kinds of liability, future
liability from this. The cost of insuring that [data] might be more than it’s worth keeping around.”
He noted that the company he currently runs insures their user data against these inevitable lawsuits,
a common albeit expensive practice. “It wouldn't be worth keeping the data around for that,” he
concluded. Some thought this idea was conspiratorial, others saw it as feasible explanation for what
happened. As Troy surmised, “It totally could be intentional. Somebody could say delete that, and
it's gone. But I think that would take a pretty bad person.”
Regardless of what happened, what comes through employees’ shock and speculation is a
major factor in decisions about sunsetting: the idea that maintaining user data has risks, while
dismantling and deleting can resolve these risks, constructing a stable albeit destructed end state for
a platform.
Vine: Pulling Apart a Platform, Assembling an Archive
Vine employees also grappled with the materiality of their declining platform, but it was
under different social and technical circumstances than those experienced by MySpace. For one,
Vine had established a sunsetting plan in which they would not delete user data. Instead, they would
shift user-generated content into a static, archived state. Moreover, Vine sunset in different technical
circumstances than MySpace. Comparing conversations with Vine employees and press pieces about
Vine with the same resources for site likes MySpace and Friendster, Vine had not experienced
technical hurdles that were comparable in magnitude to those other platforms, even as it grew
rapidly.
This did not mean that Vine was free of technical problems while it thrived or when it
declined, nor did it mean that the technical process of sunsetting was easy. As I’ll explore here,
pulling the platform apart, transforming content into an archived state, and then maintaining it in

168
this state, required significant technical investment from the Vine sunsetting team. A major part of
this technical investment was reducing the technical infrastructure that its hosting relied on. In this
section, I explore the process of sunsetting Vine technically, discussing the challenges in separating
formerly interdependent layers of the stack, how economic factors played a major role in the
material decisions that were made, and what it looks like to design a platform’s afterlife with an eye
toward durability and ease of maintenance. I draw significantly from one member of this Vine
sunsetting team–Drew–who led the technical sunset and still maintains Vine’s archive today.
Pulling Apart the Stack
“The thing is, some large systems can be architected well to where they can be pulled apart easily.
That doesn’t happen magically. You have to really think about that while you’re designing the
system,” Drew noted. “This one was not like that at all.”
Drew was beginning to describe how he went about sunsetting Vine from a technical
standpoint, previewing the interdependency of software and hardware that became readily apparent
as he chronicled this process. Employees’ goal in sunsetting Vine was to preserve accumulated user
content, and, as much as they could, to preserve this content in perpetuity. What this translated to
from a technical standpoint was changing the system first and foremost from a ‘read/write’
operation to one that was ‘read only,’ meaning that new content could only be ‘read,’ or accessed
and viewed, but not produced, or ‘written,’ to Vine’s servers and databases. Moreover, it meant
constructing Vine’s afterlife in such a way that it made it easy to maintain from the outset.
Much like how MySpace’s original architecture shaped the magnitude of commercial digital
loss and choices during sunsetting years later, Vine too was forced to wrestle with the installed base
of that platform’s earlier construction. Indeed, Drew’s remark that some systems can be dismantled
if they are architected in the right way–but that it wasn’t the case for Vine–also recalls Peter’s
characterization of sand mandalas and the engineering ethos. That is, Drew’s statement tracks with

169
the logic that Vine, like most sites, was never engineered to be taken apart, and thus grappled with
an architecture built with persistence as the assumed state. Because of this, pulling apart the
interdependent layers of the stack took significant effort on the part of the engineering team that
remained for the sunset. As Drew continued:
It took a lot of effort to unravel different systems. It’s very difficult to convey how hard that
is. You don't know what you’re going to break. You’re working on years and years of stuff
that has been layering, and layering, and then you've got to pull it apart. That took quite a bit
of effort just to rip out the recommendation engines, to rip out the whole timeline, to rip out
all that kind of stuff that we weren’t going to retain.

As can be seen here, the installed base challenged employees because of the stack’s many
entanglements and interdependent layers. Drew later noted that because of these interdependencies,
some components of Vine’s API were conscientiously maintained in the archived version even
though they no longer functioned (for instance, counting the number of times a Vine had been
played, or integrations with TV and Apple watch software), to avoid the possibility of causing errors
with old API clients.
At the same time, Vine’s installed base could have been one of the things that permitted
Vine to be so readily cut from Twitter’s holdings in the first place. As Drew explained, early on in
Vine’s tenure they had tried to integrate Vine’s technical systems with Twitter’s systems, but this
attempt to merge failed. Because of this, Vine’s architecture was “isolated away in the cloud” from
Twitter’s. While this connection is inferred, it is logical to think that if Vine and Twitter’s
architecture had been more closely linked, with interdependencies between Twitter’s chief product
(Twitter itself) and Vine, it may have been less of an easy decision to sunset the app entirely.
Because of Vine’s desire to preserve user content, its sunset was not as simple as wiping their
disks and repurposing hardware. Moreover, before Vine employees could even think about how best
to preserve user content, the first step was making sure that any systems they had were legally
compliant. Much like MySpace employees, who first considered any legal parameters when

170
deliberating on what to delete in their 2013 transition, legal issues were the first priority toward
making a sustainable archive.
Most of the work is actually in moderation tools, because you can't just put 500 million
profiles online and two billion posts and just leave it. There’s GDPR. There's Californian
data privacy rights. […] There’s also law enforcement stuff. Most of the work went into
making sure those systems worked. Then the last bit was just preserving people's profiles.

At the time of the sunset, the most known legal restriction was law enforcement’s use of data,
similar to MySpace’s era. In 2018, the Vine archive database was re-architected so that their systems
could be compliant with the European Union’s General Data Protection Regulation, or GDPR. This
included deleting all personally-identifiable metadata–whether a phone number or a personal
description in a user’s bio–but maintained a user’s screenname, their posts, and their posts’ captions.
With compliance in mind, the engineering team designed the archive as a static website.
Rather than being deployed on hundreds of industry-grade servers, the running of which cost
thousands of dollars per day to accommodate a dynamic app, all files were ported to a static, or read
only, website. The new static site was home to “billions and billions of JSON files,” Drew described,
lightweight text files that in turn referenced user-generated content stored in S3, an Amazon cloud
storage service known for being inexpensive. Indeed, as I’ll describe next, designing an afterlife for
Vine meant thinking through how infrastructure costs could be reduced such that the maintenance
of the archive would never show up as a major expense on the Twitter organization’s balance sheet,
an occurrence that could prompt the permanent removal of Vine videos.
Maintainability Means Controlling Costs
When Vine employees decided they wanted Vine’s content to live out past the six-month
window initially given to them by Twitter executives, the challenge was to design for maintainability,
or ease in keeping the system in working order over time. When I first imagined what the necessary
maintenance of a platform’s architecture would look like, and how it would be designed for long

171
term maintainability, I pictured things like reducing the complexity of the system, breaking software
interdependencies to have a simpler entity to maintain or moving to resilient hardware and
components resistant to obsolescence. These things indeed were part of the process. However, what
maintainability primarily meant in Vine’s case was making a system that was as cheap to run as possible.
As Drew noted, when Vine’s sunset was announced, “all the priorities and everything changed. It
was more: how do we keep it alive as long as we can? That basically boiled down to entirely two
things. First one was cost and second one was engineering hours.”
By focusing on controlling the cost of the archive as well as the amount of time spent
working on the archive, the goal was generally to reduce the resources dedicated to Vine. But those
two things in turn boiled down to one: because labor also represents a significant cost, what Twitter
wanted for Vine was to make it cost a negligible amount of money. Much in line with the idea of
reducing resources dedicated to maintenance by restricting the time and money spent, technology
historian Nathan Ensmenger (2014, p. 2) has noted that software maintenance represents the “most
time consuming and expensive phase of development.” As he discusses, software maintenance costs
have made up between 50% and 70% of expenditures on software development since the early
1960s. It was this phase that Twitter wanted to minimize. Because content no longer holds the
promise of economic value, maintainability in the corporate context of sunsetting means lowering
economic cost. The team that sunset Vine eventually lowered the price of maintaining the content
by a magnitude of 50.
A Stable End State
By pulling apart pieces of Vine to transform it into an archive that was easier to maintain,
Vine employees attempted to construct a stable end state for the platform. Stability–something that
engineers design for, if we return to Peter’s characterization of that ethos–is not simply the material
durability of a system, however. Stability here was generated by reducing cost, reducing engineering

172
time, reducing the hardware that it ran on, reducing the information that was saved about users, and
reducing the complexity and size of software that it was composed of. This path of reduction was
ironically a bid for presence: reducing the platform enough that it became functionally invisible for
the holding company while remaining visible for former users.
As much as the process of sunsetting attempted to reduce Vine in the eyes of Twitter, its
post-sunset presence was nonetheless a risk that Twitter was taking on. Just as Krish at MySpace
noted that maintaining data is costly and risky–providing opportunity for data breaches and
subsequent lawsuits–Vine employees had to “reduce risk so that we can justify that this thing can
live on,” as Drew noted. The prospect of security breaches “can scare people,” he continued.
Deletion is in the end the ultimate resolution to security risks. “Nothing is perfect from a security
standpoint, but that's one of the best things you can do is just get rid of as much code as possible.
For every line of code, there's a potential exploit. It's just a matter of deleting as much as you can.”
A stable end state had to be designed such that there were no surprises, no breakdowns that
suddenly thrust Vine into a state of unwanted visibility. Instead, Drew explained the design for a
stable end state almost as a mantra: “less complex, less code, less things to worry about.”
Vine the app was rebranded on the app store as Vine Camera on January 17, 2017. The
archive went online. The archive would change for GDPR compliance in 2018, and eventually its
complexity was reduced even more. In 2019, the archive was no longer browsable from the site
itself, and one could only access Vine content if they had the link to a specific Vine video or access
to a Vine that had been posted on Twitter.
January 17, 2017, was likewise the last day any Vines were made. An employee I spoke to
who had left on the day of the layoff in 2016 recalled producing one final video, the last Vine to be
made and the last item that would make it into the Vine archive. He recounted how:

173
They shut down all posting in January. I found out when they were going to shut it down,
and I talked to people who were still at the company and was like: “Can you leave it open for
a couple minutes for me to get the last Vine up?” They shut it down for everyone's user ID
except for mine, and I posted the last one.

I asked what the last post was of. He rustled up a link to it from a Google document where he had
saved URLs for individual Vine posts, the links connecting to the Vine archive. It was a video that
played to a song with the lyrics, “Young and we alive. We never gonna die” (West, 2016), the words
flashing over a grainy, descending sunset (Figure 8).

Figure 8. Eight stills from the last Vine uploaded.
From L to R. Upper: Young / We / Alive / symbols. Lower: We / Never / Gonna / Die

Conclusion: The Construction of Destruction

This chapter has shown how sunsetting proceeds from a material and technical standpoint.
In doing so, it has observed how early forms of technical architecture come to influence later forms
of sunsetting, whether MySpace’s technical debt-ridden infrastructure informing the construction of
a separate site as it transitioned in 2013 or Vine’s cloud-based architecture informing how the site

174
was pulled apart and the archive was constructed in 2016. This need to grapple with an ‘installed
base’ of earlier technical structures shows how sunsetting is essentially being shaped even as a platform
comes into being. Peter from Friendster and Drew from Vine noted that sites could be architected in
such a way that they could be pulled apart or closed easily when they needed to be. But the fact that
most platforms are not architected in this way does not mean that earlier design processes are
disconnected from later processes of closure.
Just as the ‘social factors’ explored in Chapter 2 and 3 were never separable from material
forces, this chapter also shows how even when explicitly focusing on materiality and technicity, it is
impossible to separate these factors from their social context. From the way that MySpace’s ‘churn
and burn’ ethos shaped their installed base to the way that employee values shaped the decision to
preserve limited content in their transition to the new site, personal values and worldviews
influenced how the material process of sunsetting happened. The same can be said for Vine, who
chose a different path according to values expressed by their employees. Moreover, these cases show
the significant interaction between socio-economic forms of valuation and perceptions of a
platform’s materiality. That platform infrastructure cost money made it especially visible across all
levels of a platform organization in moments of decline and sunsetting, a material visibility that
differed from earlier, ‘successful’ eras.
The construction of destruction indeed begins in early technical design and continues
through the tenure of a platform. However, the stages explored here–where data is actively deleted,
and a platform’s layered stacks are disassembled or rearchitected–are particularly acute moments in
shaping a platform’s afterlife and the magnitude of commercial digital loss. These stages show
employees’ attempts to construct a stable afterlife for a platform from varied components, attempts
that took different forms for MySpace and Vine. A stable afterlife for Vine meant rearranging
platform components into a pared down, read-only experience of previously created content, one

175
that was cheap, easily maintainable, and resistant to security breaches. For MySpace, employees
attempted to create a stable whole by deleting the previous platform and most of its user content,
removing a form of technical debt that they believed impinged on the new platform’s ability to
succeed. Stability–what the dominant engineering ethos that started this chapter is defined by–is
therefore not purely material. Moreover, stability in the case of sunsetting does not necessarily refer
to an immovable material structure. Rather, stability in sunsetting often means the construction of
absence and loss.

176
Chapter 5

Sunsetting, a Story: Media Accounts of Platform Closure

The shock had not worn off for employees after Vine’s sudden internal announcement when
press accounts about the platform’s closure–and a broader public response to it–started rolling in.
That same day, the NYTimes ran their story “Vine Is Closing Down, and the Internet Can’t Stand
It” (Rogers, 2016). TechCrunch–a site that one interviewee described as “the Wall Street Journal of
the tech industry”–ran a story too (Perez & Roof, 2016). The stories kept on coming in the ensuing
days: more eulogies in the New York Times (Hughes, 2016), WIRED (Raftery, 2016), the
Washington Post (Mudallal, 2016), Engadget (Low, 2016), NPR (“Twitter to Discontinue,” 2016),
and even a bit on The Daily Show with Trevor Noah (2016). Allison, who helped craft the initial public
announcement of its closure, noted that Vine never seemed to stop being talked about, whether in
press accounts or by former users: “That has to mean something,” she said. “I don't know what it
means, but that has to mean something.”
As in death, so in life. Press coverage had consistently been used as a storyteller for
employees to make sense of the platforms they worked for. Press was a barometer for the cultural
relevance, business health, and competitive landscape of each platform in each era. As one
Friendster employee would describe to encapsulate the overnight success of their site: “You had
[founder] Jonathan Abrams on Rolling Stone. We had MTV. We had all the hottest things in media at
that time knocking on Friendster's door.” The feeling was like that expressed by employees at other
sites. Brendan had been hired during MySpace’s peak years. Arriving to their Beverly Hills
headquarters for his first day after being hired away from his post-grad job in Indiana, MySpace’s
cache was clear from the magazine covers that adorned their lobby. “I came in, and to see the big
MySpace logo in the reception area, and all these magazines that I was familiar with like Fast
Company, and Fortune, and Time have stories about MySpace,” he noted, clearly illustrated the

177
significance of the company he had just joined. When a platform sunset, press coverage was a means
of making sense of what had happened–one version of a story employees had lived, narrated back to
them. “Vine was my first time being on the inside,” Austin, the platform’s music coordinator, noted.
“You can read the press and be like: nope, nope, nope, yeah, yeah, nope, nope, nope. You can figure
out who got it right and wrong.”
Press coverage, however, does more than give a synopsis of the reasons for closure to
employees and to the broader public. Rather, these synopses are in themselves bearers of more
substantial meaning. As I show in this chapter, they reflect broader ideologies about technology,
express latent expectations around digital memory infrastructures, and reveal important themes
about platforms’ continuing place in public life, even in death.
More specifically, accounts of platform sunsetting are public narratives of digital memory
infrastructure when it breaks. In previous chapters I have established that digital memory
infrastructures’ afterlives–their brokenness, their ruin, their data absences–are constructed and value-
laden just as much as functioning, persistent, and ‘successful’ technological systems are. Analyzing
press coverage of sunsetting therefore focuses on the public nature of infrastructural breakage, and
engages with the adage first put forth by Star (1996) that infrastructural systems are invisible until
broken. This moment of breaking presents an opportunity, then, to understand otherwise latent
public assumptions around data permanence, memory practice and platforms. In what follows, I use
theories from media studies and STS to analyze press accounts of sunsetting centered around the
three primary platforms analyzed in this project: Friendster, MySpace, and Vine, as well as
GeoCities, a significant platform that began in 1994, was purchased by Yahoo! in 1999, and closed in
2009, deleting all user content (some content was preserved by the Internet Archive, which I discuss
in the Conclusion). I include GeoCities here because it has similar cultural significance to the other
three platforms I analyze in this project, with similar amounts of press about its sunsetting. In total,

178
the analyzed corpus consists of 252 media items, the breakdown of which can be seen in Table 2.
Through these narratives, I show how public expectations of digital memory infrastructure
as stable hosts for user-generated content belie the practices of sunsetting shown in preceding
chapters, just as I demonstrate how ideologies of innovation and disruption help position platform
sunsetting and commercial digital loss as natural and immutable processes. I observe the unresolved
tension between these two perspectives–between expectations of stability and expectations of
disruption. Finally, I conclude by showing how these accounts, when addressing the cultural effects
of digital loss, rarely consider social difference in their treatment of this issue. I thus draw attention
to the fact that it is not only digital loss, but also digital preservation, that can threaten to re-entrench
representational harms.

Table 2 Collected Media Accounts of Platform Closure

Platform name No. media items Collection event

GeoCities

64

2009 closure
23

Friendster

66

2011 gaming site transition

MySpace

41

2019 loss of content

Vine

81

2017 closure

23
This includes the primary GeoCities portal. It does not include GeoCities’ separate Japanese subsidiary,
which ultimately shuttered in 2019 (Steger, 2018).

179
Biographies of Breakage

Social media platforms increasingly operate as infrastructures (Helmond et al, 2019; Plantin
et al 2016; Plantin, 2018), in part because they reproduce infrastructural characteristics like
embeddedness and invisibility (Star 1996). First conceptualized by Susan Leigh Star in “Ethnography
of Infrastructure” (1996), the invisibility characteristic–expressed in the idea that infrastructures only
become visible when broken–is a hallmark characteristic in STS approaches to infrastructure
(Larkin, 2013).
The ways this brokenness-visibility paradigm has been used as a blanket description for all
infrastructure since then has been examined by the anthropologist Brian Larkin (2013), who argues
that infrastructures are in fact strategically visible, even when functioning. Because people see,
interpret, use, and give meaning to technological systems in varying ways (Mackenzie, 1996),
infrastructures are likewise “semiotic and aesthetic vehicles oriented to addressees,” acting not only
as substrates for modern life but as performative and public material agents (Larkin, 2013: 328).
The stories that become attached to infrastructures—or any technological system–are
ideologically significant (Crawford, 2007; Natale, 2016). STS literature has in fact long proved this
out, whether in theories of technological determinism—which shows the ways beliefs about
technological change are significantly linked to ideas of social progress in the United States and
elsewhere (Smith, 1994)—or through writing about sociotechnical imaginaries, which has shown
how collective beliefs about technology are mobilized in shaping desired societal futures (Jasanoff &
Kim, 2009).
The stories attached to media technologies, and their lifecycles, are likewise ideologically
significant.
24
The media discourse around platform sunsetting that I’ll consider in this chapter are

24
Of course, the entire notion of a media lifecycle is itself contingent and resonates with the media biography
trope discussed here and by Natale (2016). See Conclusion for extended discussion.

180
what media studies scholar Simone Natale would call “biographies of media” (2016, p. 436) or those
“recurring and highly standardized tales” drawing on experiential biographical events like birth,
aging, and death, which “appeal to our taste for narration and remain in our memory with an
effectiveness that perhaps only storytelling allows” (p. 436). These tropes about media’s emergence,
development, decline, allow social groups to make sense of media change, and of novel experiences,
social configurations, and events that may be precipitated by this media change (ibid).
If it is possible for technological systems to be freighted with meaning, what do they mean
when they are broken? How do they act as semiotic vehicles then? Ballatore and Natale (2016) have
explored this question in the context of the emergence of ebooks and contemporaneous narratives
around the so-called “death of the book.” As they note, the death of a particular medium is a
longstanding and recurring myth that arises when a novel media technology emerges. The idea,
however inaccurate in practice, that one medium replaces and ultimately destroys a related medium
not only helps people make sense of shifting sociotechnical configurations (ibid; see also Marvin,
1988), but ultimately strengthens broader social beliefs about the transformative nature of
technological innovation (see also Lepore, 2014).
How do narratives about sunsetting interface with existing biographical tropes, especially
around a medium’s replacement, death, or destruction? While there is limited scholarly literature on
technologized ends, especially as they apply to the platforms of today, failure and death have been
strategically framed since the popular advent of social media platforms in the early 2000s. In the
wake of the dot-com crash, that sudden downturn of the Internet economy in 1999-2001,
technology industry gurus like Tim O’Reilly pushed a new narrative for the development of the
online world known as Web 2.0. Web 2.0 narratives heralded a new online era dominated by
platforms like Friendster, MySpace, and Facebook. This versioning attempted to communicate to

181
both skittish investors and a skeptical public that this new cache of web-based startups had learned
from the failures, and death, of the retrospectively named ‘Web 1.0’ (Ankerson, 2018).
Moreover, narratives about failure and death hold significant ideological weight within
Silicon Valley and related techno-cultures. Media scholar Nora Draper (2017: 1) has noted that “‘fail
fast, fail often’ is the unofficial mantra of Silicon Valley,” reflecting a broader neoliberal ideology
“that encourages an unwavering focus on staying ahead of the curve and discovering the next big
thing.” In an article that draws on my interviews with Friendster employees, I have shown how the
closure of a social platform is often phrased in terms of its ‘death,’ and argued that employees
mobilize stories about a platform’s ‘death’ as instructive assets within their Silicon Valley-based
careers (Corry, 2021).
While this scholarship has provided a foundation for understanding failure and death’s
significance for actors within Silicon Valley, it leaves room for asking about the ways that failure and
death are understood from the outside, whether users who interact with platforms or media who
observe this industry’s broader patterns. Indeed, press coverage is one way of understanding these
narratives’ ideological positioning, as scholars of media discourse have argued that news media
encodes ideologies that both shape and reflect their broader social contexts (Cotter, 2015). This
leads to questions I’ll address in the analysis of sunsetting press accounts: What are the stories that
are told about digital memory infrastructure’s breakage? Or, more concretely, how is platform
sunsetting narrated?
Infrastructural Expectations

In Chapter 1, I argued for a reconsideration of what seems to be social media platform’s
latent functions: perpetual content storage and distribution. While scholarship focusing on social
media platforms’ temporality has largely highlighted the way these sites emphasize ‘liveness,’ (Kaun
& Stiernstedt, 2014; Mosco, 2014), or the built-in favoring of new information, I noted that the

182
consistent ability to store and serve up content is a definitional part of what it means to function as a
platform (Gillespie, 2010; 2018). The seemingly perpetual accessibility of previously posted content
is a crucial part of what sets social media platforms apart from, say, instant messaging systems.
Rather than thinking of the archiving functions of social media as obfuscated primarily because of
an emphasis on liveness, however, we might also think of the archiving functions as largely invisible
because they are so endemic to the functioning of social media that they are unremarkable,
backgrounded, banal.
Larkin’s (2013) treatment of the infrastructural invisibility paradigm has asked analysts to
think broadly about the differential and strategic visibilities of infrastructures, just as other studies
have pointed out that the seamless and invisible functioning of infrastructure is a privilege of
wealthy communities largely situated in the global north (Anand, Gupta, & Appel, 2018; Edwards,
2019; Graham and Martin, 2001). Regardless, as Paul Edwards (2003) and Ashley Carse (2016) have
argued, seamless functioning, if not invisibility, is a common infrastructural goal. The social media
platforms I examine here are indeed situated in some of the wealthiest communities in the global
North–based in the major US metropolises of San Francisco, New York, and Los Angeles–and have
largely imagined their users (however inaccurately) as functioning within the infrastructural norms of
these communities (Corry, 2021). As such, these sites position themselves alongside infrastructural
systems that do work to become invisible substrates supporting a wide range of user behaviors,
whether social, economic, educational, or cultural in nature (Plantin et al, 2016; Van Dijck et al,
2018). In their invisible and unremarkable general functioning, platforms not only appear as
infrastructural because they support the functioning of modern life (Carse, 2016), and, as I’ve
argued, the functioning of modern memory, but because they are implicitly expected to work for
these memory purposes–that is, to preserve data for the long-term.

183
Sunsetting and commercial digital loss are violations, then, of expectations around the
perpetual nature of content accessibility, and function within the brokenness-visibility paradigm first
put forth by Star (1996). Media narratives of sunsetting are therefore generative moments to parse
sociotechnical expectations around these systems, and particularly, about these systems’ relationship
to the accessibility of the past. In turn, this analysis asks about the expectations of digital memory
infrastructures expressed in these accounts, just as it examines the broader cultural discourses
around content permanence and ephemerality. Using a grounded approach, in which themes
emerged from multiple rounds of in-vivo analysis, in what follows I explore media narratives during
and following the sunsetting of GeoCities, Friendster, MySpace, and Vine.
25

26

Narrating Sunsetting

Despite their ubiquity in the broader history of platforms, long drawn-out decline and
eventual closure are not necessarily common topics for journalists covering the tech industry. These
platform organizations, after all, are typically on their last legs, and have been deemed to have lost
their cultural relevance–and therefore their newsworthiness. There is no obvious template, then, for
reporting on these occurrences, outside of those that exist for any form of reportage. Press
narratives of sunsetting are therefore marked by complex patterns of similarity and difference,
possessing common themes but also a diversity of narrative styles and a range of metaphor and
vocabulary used to get at just what sunsetting means, and why it matters.
The simplest of these accounts were brief and found in mainstream news sources rather than
technology-focused outlets. They announced the date of closure, typically included a brief statement
by the network, and frequently mentioned their plan for deleting or preserving the data. Even in the

25
For additional information on the methods used in this section, refer to Appendix A.
26
It should be noted that press coverage of these closings span a decade–from GeoCities’ 2009 closure to
MySpace’s 2019 data loss–and therefore represent different eras in social media history. Nonetheless, the
narratives that emerged were relatively consistent through these accounts.

184
most limited of accounts, however, these stories always noted a reason for the closure. Perhaps this is
unsurprising: asking why–along with the who, the what, the when, and so on–is a basic tenet of
communicating about an event of note. Yet what was remarkable across these accounts were the
variation in reasons chosen for why a closure happened when, in essence, platform closures are
iterations on the same theme: the movement of users away from a social network, the attendant loss
in revenue or hope for revenue, and the comparative cost in running something that seemed to only
be a financial burden for a holding company.
Across business literature–particularly investment blogs or finance-focused outlets like the
Wall Street Journal–narratives of closure often cited the reason for closing as rooted in cost-cutting
measures or suffering stocks. Announcements of GeoCities’ sunsetting noted that “the trimming is
part of a process that started in 2007 while Jerry Yang was still chief executive, to close down
services that aren’t profitable or don’t fit into company’s long-term vision. (‘Yahoo to Pull the Plug,’
2009) while those chronicling Vine’s demise described it as a side-effect of “Twitter's quest to
become profitable” (Cochran, 2016). Because most of these platforms were at the time of closure
owned by parent companies, these narratives were also careful to note how the parent company
fared after these austerity measures. Twitter, cutting Vine, had “revenue [climb] more than analysts
expected as the struggling social media company announced 350 job cuts, or about 9% of its
workforce” (Molina & Guynn, 2016). Accounts noted that Yahoo’s restructuring, which included
removing GeoCities and cutting “thousands of jobs,” “seems to have worked a bit for Yahoo! as it
turned a profit of $ 187 million this quarter” (‘The Internet’s First,’ 2009). In these stories, platforms
are valued as financial assets, and their eventual removal is a logical function of corporate
vicissitudes.
Other financially oriented takes noted that closures were caused by internal failures on the
part of platform itself, rooted either in an inability to monetize and prove their financial value or

185
their failure to maintain growth and therefore illustrate their future financial value. A Wall Street
Journal article discussing a wide range of platform closures chronicled the failure to maintain user
growth as the root cause in numerous cases, noting that sites from Friendster to YikYak “plateaued”
“stagnated” or “stalled” (Brown, 2017). The inability to monetize was the culprit that some chose,
noting that platforms–especially Vine–couldn’t prove their worth when compared to the cost of
running them. “Video is expensive,” an article in The Globe and Mail would note. “And the six-
second loops are hard to monetize, in part because Twitter itself has been slow to invest in or
exploit the platform” (Erin, 2013).
As these accounts demonstrate, and as is true in the case of employee discussions of
sunsetting (Corry, 2021), there are a multitude of reasons that one can discuss as the root of a
platform’s ultimate demise, a litany of lines that can be drawn between causes and effect. However,
more than any of these reasons that have been explored so far–whether financial or growth-
oriented–accounts of sunsetting chose to position closure within broader stories about competition
between social platforms. This came out in what could be called ‘the list,’ a chronological itemization
of dominant social networks. What ‘the list’ looks like, and what it means for broader beliefs around
technological change is discussed in the next section.
The List

One of the first articles about Vine’s closure came out in the New York Times just hours
after the company announced to its employees–and then to the greater public–its impending
sunsetting. Written by their technology reporter Katie Rogers, the lead paragraph read (2016):
If you picture social video as a colorful ecosystem full of emoji and chat bubbles, you really
should try to think of it more as a jungle -- a place where the strong eat the weak. Periscope
ate Meerkat. Instagram's new video features might be chewing on Snapchat. And Facebook
Live is trying to gobble up everything in sight. The latest casualty is Vine, a six-second video
app owned by Twitter.

186
Rogers’ account is emblematic of stories of sunsetting. It chronicles the demise-du-jour of a
platform–in this case, Vine–and positions its closure within a competitive landscape of disruption
and displacement. As noted, while some accounts chose financial root causes to discuss platform
closure, the more common way to describe the back story of a platform’s sunsetting was through a
progressive list, almost a mini history of platforms that successively disrupted and replaced one
another. Another account would describe (Jamison, 2018):
Friendster was the first social networking site to reach one million users, but it fell to
MySpace, which fell to Facebook. News aggregator Digg lost so much ground to Facebook
and Reddit that it had to be broken up. Being any other type of platform business - a
business whose customers provide much of the value - is hard, too: Yahoo! was overtaken
by Google. In the online retail space, NetMarket preceded Amazon, but Amazon has proved
to be superior.

By enumerating this list, wherein one platform inevitably replaced another, these
descriptions attest to an evolutionary view of technological development. Like in Roger’s account
(2016), in which the “strong eat the weak,” sunsetting stories are often explicitly framed in a
Darwinian view of technological change, governed by the naturalized rule of the “survival of the
fittest.” This narrative resonates with cultural tropes around media change (Ballatore & Natale,
2016), which conceptualize novel media forms as emerging to then compete, replace, and ultimately
destroy existing media forms.
This trope of rapid medium obsolescence caused by the introduction of a new medium is
closely related to theories of technological change centered around “disruption” and “innovation.”
As technological historians have chronicled, notions of disruption and innovation became popular in
the US in the 1980s: optimism around the terminology of progress waned in the wake of the
Vietnam War, anxiety grew around the loss of an American empire, and US politicians became wary
of non-Western nations’ growing economies based in part on microelectronics manufacture
(Castells, 2010; Lepore, 2014; Vinsel 2014a, 2014b). Disruption, which describes the use of new

187
technology to offer alternates to products sold by established companies (Lepore, 2014), and
innovation, a slippery term signaling either betterment through novel technology or simply
transformation through these technologies, seemed to promise to shake the US out of economic
stagnation while allowing it to regain its footing on the global stage. Pulled from business literature
developed throughout the 20
th
century, these theories have been applied widely, from education to
journalism to the social world at large. While inherently promising perspectives that were fresh and
new, these terms in fact recreated longstanding cultural beliefs about technological development
equating to social progress, prominent American ideals since the mid 19
th
century (Marx, 1997/2010)
These buzzwords–still popular forty years later and especially so in Silicon Valley techno-
cultures (Vinsel, 2014a; 2014b)–represent what historian Jill Lepore has characterized as an historical
ideology (2014), a theory of change organizing the past and appearing to predict the future.
Ideologies of disruption and innovation in turn position failures, shutdown, and yes, sunsetting, as
inevitable, natural, and often healthy processes applicable across contexts. Whether describing the
organization of the economy, of social life, or of the past, these processes retrospectively position
certain failures as pretext for the more efficient, forward-thinking, perfected predecessor.
The use of evolutionary metaphors ensconced in the narrative organization of ‘the list’
indeed attest to a belief that sites have natural life cycles. As a pundit in one account noted: “‘These
Internet businesses tend to have a cycle,’ said Mr. Wolf, the former MTV president who once
sought to buy Facebook for Viacom. ‘There’s a lot of people who wonder if the same thing will
happen to Facebook’” (Arango, 2011). Implicit life cycles were also communicated through accounts
that discussed the inevitability of closure under circumstances where other companies have offered
seemingly superior features: “Moreover, with Instagram and Snapchat, where users can post
interesting and longer clips, with a plethora of other options, Vine was bound to fall” (“Twitter Kills
Vine,” 2016). Sunsetting’s apparent inevitability and cyclical nature in these accounts paint this

188
process as a natural one within an already naturalized ecosystem of disruption and innovation. By
fitting neatly within these tropes, sunsetting accounts reflect and reinforce innovation and disruption
as historical ideologies, blanket truths describing both technological and social change.
While disruption and innovation are often depicted as self-reinforcing prophecies, the
particular flavor of disruption and innovation were up for debate. Did platforms sunset because of
competition and disruption–as new companies offered superior alternatives? Or was it the attraction
of innovation–the cultural cache of the new enticing users to join up with one and drop out of
another? As one account argued, the pattern of entrance, replacement, and closure embodied in ‘the
list’ occur because the sites that succeed are constantly working to disrupt the norms set up by their
competition. In this way, these accounts argue, social networking itself is constantly improved
(Wilson, 2011):
To save time, let's pretend Friendster came first, with its ability to reconnect with old friends.
Then MySpace came around, and added musicians. Then it was Facebook, which started as a
college thing but branched out. And you can't forget Twitter. Or LinkedIn. Or Tagged. Or
MyYearbook. Or ... you know what? Here's a blank space so you can write in the one I
forgot _______________________. New social network, same basic story. The new
network comes to the gate. Backers of the current leader scoff, "How does your site improve
networking?" Heck, the same question gets asked back to the ones that are established. The
ones that can answer the question stay; the ones that can't get ignored, shut down or sold
off, with somebody taking a huge loss.

Other versions of ‘the list’ describe social media users as fickle, coming and going in generations,
attracted to the newest entrant regardless of its features (Dumenco, 2012).
My point is that the lesson of the “Attention Economy” is that we all get sick and tired of
everything--so much so that the epic tech upstart-vs.-incumbent narratives of yore now seem
not so much like business-school case studies, but universal memento mori. Remember
when AOL beat CompuServe? And Yahoo beat, what was it, Lycos? And MySpace beat
Friendster?

These takes position platform closure as byproducts of natural cycles of replacement but place these
cycles within broader cultural patterns privileging the ‘new and cool’–or the cool because it’s new.

189
Analyst Jeffrey Cole, in one article, would compare sites like MySpace to physical spaces whose
value lies in their clientele and cultural cachet. “To a teenager, an online community is like a
nightclub. And when the nightclub becomes too popular or the uncool kid starts showing up, you’re
out of there” (“Dow Up,” 2010).
Yet in describing–whether implicitly or explicitly–the naturalized progression and inevitable
life cycles of platforms in ‘the list,’ these accounts inexorably bump up against that network which
seems to defy these cycles: Facebook.
Where the List Ends, Infrastructure Begins?: Facebook

In theory, the naturalized cycles of disruption and innovation in which sunsetting is
understood as an inevitable stage should continue unabated. These are cycles, after all: repeated
sequences of events, more “universal memento mori,” as the earlier account noted, than unique case
study (Dumenco, 2012). At the same time, the list, that mini history of social platforms, typically
ends with Facebook, or Google, and their attendant products’ dominance. A tension arises at the
end of the list, then: between these platforms’ dominance and the repeated cycles that these
accounts otherwise described.
27
For instance, referring to the closure of Friendster, Ivor Tossell in
The Globe and Mail writes (2011):
This could never happen to Facebook … right? Predicting Facebook's demise is always a fun
sport, even if there are no signs that indicate it will suffer the same fate as Friendster,
MySpace and the other social networks it crushed.

And Iain Tait in the British magazine Campaign writes (2011):

27
These were tensions that interviewees also wrestled with. As Michael, a cofounder of the mobile platform
Couple, noted: “Evidence shows that there is a lifecycle. Facebook is a weird one. I imagine Facebook as an
entity will continue, probably be there a long time. And they'll just keep kind of rolling with the times, and
just like acquiring, building.” At the same time, others project Facebook’s demise, believing it to have
“stagnated,” as interviewees like Fidelia at Friendster described.

190
We've seen mass online migration in the past. We watched as MySpace hollowed out
Friendster. As Napster shut down, people found Limewire, Kazaa and Soulseek. And
recently, Facebook has left MySpace gasping for air. But Facebook is different, right?

These accounts grapple with a contradiction: does Facebook represent the end to these life cycles,
having become so “fit” for the platform environment that it stabilizes and remains dominant? Or
have we just not reached the moment at which it too, like Friendster, MySpace, GeoCities and Vine
before it, atrophies, fades, and sunsets?
Another way to think about this tension, a rather unresolved one in these pieces, is that these
articles are contending with the evolving infrastructural characteristics of platforms, most notably,
the evolving infrastructural role of Facebook. What might otherwise be described as “fitness” for
the platform environment, successfully propagating itself into the future while squeezing other
platforms out, has been alternately described by platform scholars as a move toward consolidation
and infrastructuralization. As Gillespie (2018, p. 14) and other platform studies scholars have argued
(Helmond et al, 2019; Van Dijck et al, 2018), online interaction is increasingly consolidated, as
“encounters with information and people that were once scattered across the web have been largely
gathered up by a small set of companies onto a handful of social media platforms,” Facebook
among them. This funneling toward a few dominant and for now persistent platforms has been
accompanied by the increasing infrastructuralization of these platforms, meaning that they
reproduce characteristics commonly associated with infrastructure (e.g. Star & Ruhleder, 1996), but
also that they appear to provide “the stable foundation of modern social worlds” (Edwards, 2003, p.
186).
Indeed, this seemingly infrastructural stability of platforms like Facebook is at the root of
this tension explored in these accounts. These accounts ask: what are we supposed to make of
something that acts infrastructurally, but seems part of a pattern of impermanence? There is
cognitive dissonance between what is expected of infrastructure in these contexts–its stability, its

191
embeddedness, and its permanence–and what is otherwise described as subject to the “historical
ideologies” of innovation, disruption, and replacement. I discuss this dissonance in greater detail in
the next section, exploring the infrastructural expectations around permanency and ephemerality of
user content in relation to digital memory infrastructure.
Permanence, Ephemerality, and Expectations of Digital Memory Infrastructure

The loss of user content disrupts one of the important means by which platforms support
contemporary remembering. That is, one of the ways that digital memory infrastructures support
reflection is through stable storage. Storage, as I have noted, does not constitute memory (Chun,
2011). Practices of memory, instead, are mutually shaped by everyday social cultural contexts as well
as (socio)technical factors of their form of mediation (Van Dijck, 2007). In this process, the storage
and organization of user content are crucial, non-neutral, yet backgrounded components in how
remembering and forgetting are constructed using platforms (De Kosnik, 2016; Villa-Nicholas,
2019).
The previous section highlighted the ways that these accounts grapple with the
infrastructural characteristics of platforms while placing them within broader narratives of
technological displacement and disruption. Here, I show how these pieces grapple with a particular
infrastructural feature of digital memory infrastructure–storage–and what happens when this feature
breaks, as it does in cases of sunsetting. In doing so, this analysis highlights the expectations and
norms that are entangled with digital memory infrastructure, those beliefs about how infrastructures
do function and how these systems should function.
Like the tension between infrastructure and disruption described in the last section, these
accounts contain conflicting expectations about the permanence and ephemerality of content on
platforms. These pieces communicate two contradictory expectations: that history has shown the

192
internet to be impermanent, and that sunsetting is therefore unsurprising; or that platforms are
marked by archival permanence, and therefore sunsetting is unexpected.
For instance, one elegy to Vine noted (Andor Brodeur, 2016):
On an Internet under siege by dangerous-stupid, Vine provided critical sanctuary for silly-
stupid, a safe space for playfulness and invention, and a supply of quick laughs that I'd just
assumed, by nature, was endless.

While another article argued that Vine’s closure (Swartz, 2016):

…should come as no surprise really. The tech industry is littered with dozens of mergers
that looked good on paper but didn't translate to real-world markets. Tech mergers often
flop. The concept of melding two large, disparate organizations and cultures flawlessly is
akin to adding a third leg. Overhead costs and logistical nightmares further diminish chances
for success.

Similarly, in accounts of GeoCities closure, one account noted that users had expectations of the
data’s permanence (de Vise, 2011):
All of this runs counter to the notion that anything posted online, particularly if it is
unflattering, is permanent.

While others noted that closures weren’t a surprise (El Akkad, 2013):

None of this is particularly unexpected, however. And in a sense, having a favourite Web site
or app is, over time, not that much different than having a favourite athlete.

Most often, though, closures–and press about them–are conceived of as acting as a lesson for users.
These accounts position sunsetting as a teachable public moment, an event that undoes expectations
that the internet is “forever.” I include numerous examples to show how widespread this sentiment
was:
People are always cautioning that the embarrassing tidbits we post to the Internet might
haunt us forever. But keeping vast quantities of data online isn't free: Data centres are
expensive things to maintain. Indeed, a business doesn't have to fail outright for data to
vanish; all it takes is a change of business plans. (Tossell, 2011)

Turns out, the Internet is not forever […] The bottom line is: The old adage that anything
posted to the Internet will last forever is a myth. (Sweeney, 2019)

193

Computer users who think their Flickr photos and Facebook updates last forever could be in
for a shock. The average life of an Internet page is about 100 days. (de Vise, 2011)

It was naive of me to expect a thing built on six seconds to last forever, but Vines have
grown into such a ubiquitous feature of the online landscape, it's odd to imagine the Internet
without them. (Andor Brodeur, 2016)

In other words: the infrastructural expectation of platforms is that the structure and storage of user
data is stable. But if attention is drawn to platform closure and data ephemerality through accounts
like this–and has been from GeoCities closure in 2009 through MySpace’s data loss in 2019–why do
these narratives continue to debunk beliefs? Shouldn’t they already have been debunked? On the
one hand, this sentiment–that the public should know by now these platforms are subject to closure
and data loss–are represented by those articles that argue for the “unsurprising” nature of sunsetting.
On the other hand, articles chronicling sunsetting and data loss at all are anomalies within broader
press coverage of platforms. While I assemble these stories based on their relation to closure, one
must place them within dominant discourses that attest to the permanence of the internet. For each
one of these articles there are innumerable more that recount stories of persistent circulation and
permanence–whether scandalous stories about someone’s decade-old Tweets being resurfaced (e.g.
Goodin, 2021) or subtle narratives about using Instagram to reflect on a vacation well-spent
(Compton, 2020). That is, there are innumerable more that attest to infrastructural norms of
permanence and accessibility.
When do Platform Permanence and Ephemerality Matter?

Accounts that focus on the ‘cultural’ aspects of platform closure, rather than the immediate
economic effects for a holding company, tended to offer a rationale for why the permanence or
ephemerality of user-generated content matters. Data preservation is positioned as mattering to
personal or familial reflection, or as a historical record for generations many years in the future.
However, who it matters for was rarely addressed.

194
When dwelling on the loss of content, many of these accounts liken platforms to those
physical sites that are associated with mediated modes of memory collection, recalling, for instance,
MySpace employee Gavin’s notion of content deletion as like “burning the shoebox.” As a New
York Times journalist wrote (Wortham, 2011), while chronicling Friendster’s content deletion:
But some say Friendster has unexpectedly turned into a time capsule with snapshots of who
they once were. It is a version of their history that is not in a scrapbook or dusty shoebox
but is live on the Web —for now.

Scrapbooks, as in this example, were the most common analogy for platform’s organization of
mnemonic content. “Social-networking and photo-sharing sites [have] become the personal
scrapbooks of our time” (Kang, 2011), a reporter in the Washington Post wrote, while a journalist
from The Daily Telegraph described MySpace as “the scrapbook, the fan club and the bootleg tape
collection” of its era (‘How we forgot about,’ 2019). These were among eleven others that used the
same metaphor. This is presumably because, like platforms, scrapbooks remediate and consolidate
‘scraps’ of other media–whether photographs or newspaper clippings, letters or squares of fabric–
and they tend to be added to over time.
Sometimes the person who content loss affected was phrased as the individual reader–your
ability to look back, your photos, your messages, your memories. Other times, these articles noted that
content loss during sunsetting will impact not the individual that the content chronicles, but their
future generations of relations. Harkening back to memory theorist Maurice Halbwach’s
(1925/1992) notion that elders often take on the social role of being memory keepers through their
manipulation of media, a journalist in The Observer wrote that (Tucker, 2019):
The Marie Kondo idea that you should be storing all your books, photographs and music in
the cloud, so we have nice clean shelves, is great. But just be aware that your grandchildren
might know nothing about you–unless someone is taking the time to think: that platform is
becoming obsolete, let’s make sure we download an archive.

195
If not phrased in terms of personal or familial memory, these accounts note that permanence and
ephemerality matter for the historical record. They typically draw parallels between other documents
that were designed to be ephemeral and assumed to be worthless, and the ways in which their
incidental preservation eventually became instrumental in crafting an understanding of the past. An
article in the Australian Sun Herald that addresses GeoCities closure (Turney, 2014), for instance,
compared social media production today to ephemera from the Victorian era:
It might not seem worth the bother when every tweet, Facebook update and chat avatar is as
mundane as the next, but theatre owners thought the same thing in Victorian-era England
when they issued payslips to actors. Today a program called Nineteenth Century Collections
Online has copies of them, and they give us some of the few surviving artefacts and insights
into life at the time.

When articles address these sites’ importance for leaving a historical record, they tend to draw on
the expertise of two organizations–the Archive Team (mentioned eight times) and the Internet
Archive (mentioned 16)–and leaders associated with those organizations, especially the Archive
Team’s founder Jason Scott or the Internet Archive’s founder Brewster Kahle. These are indeed the
sites and groups who often save vast swaths of data when a platform sunsets, including for sites like
GeoCities.
Interestingly, but perhaps not surprisingly, just one account out of the 252 that I analyzed
discussed how the permanence or ephemerality of user-generated content may affect users in
uneven ways. In this way, the accounts that address how social media platform data may be used for
historical reflection tend to reflect the ethos of mainstream preservation efforts from the
organizations I cite above, that envision collection through a notion of historical universality: that
what is collected is equally important to and is shared by all people (De Kosnik, 2016). (Indeed, the
Internet Archive’s slogan is ‘Universal Access to All Knowledge.’) I conclude then with the sole
perspective that asked how collection at a platform’s end might consider dynamics of power and

196
diversity in preservation. Journalist Rachael Krishna (2019), writing in The Independent after MySpace’s
2019 data loss, notes how:
When you look at what stays online and what is removed, it becomes unavoidably political;
those with money and influence remain on the front pages on the internet, deciding who
gets to sit there with them and determining the value of others’ work. In the same way that
people push for diversity of creation, people should also push for diversity of preservation.
The internet has been a powerful tool for marginalised voices: let’s make sure it isn’t easy for
those who traditionally write history to quickly take that platform away.

Conclusion: For Whom Does Sunsetting Matter?

While this project has so far focused on production-level practices in sunsetting, this chapter
reflects on the construction of destruction by analyzing public narratives of sunsetting, those stories
of digital memory infrastructure when it breaks. Through these accounts, broader ideologies and
tensions about technology have been revealed, just as implicit perspectives over who sunsetting
serves (or doesn’t serve) have been made visible.
As these accounts show, press about platform sunsetting recreates time-worn narratives of
disruption and replacement that exist for both media forms (Ballatore & Natale, 2016) as well as
technologies (Lepore, 2014). These ideologies are animated through naturalizing rhetorical devices
like ‘the list’ or the ‘lifecycle.’ These devices are naturalizing in that they place sunsetting within
theories of the natural world: through ‘the list,’ platforms are likened to organisms (sentient or not)
competing as part of natural selection, while the idea of a lifecycle for platforms harkens biological
lifecycles, with known and inevitable stages of birth and growth and reproduction and death. By
affiliating sunsetting with naturalizing metaphors, these narratives frame this process as part of
seemingly inevitable, absolute processes beyond the scope of human control–thus suggesting that
commercial digital loss is natural and immutable, rather than constructed and alterable, as I’ve
shown in previous chapters. In doing so, they enact the “aura of phantom objectivity” that Leo

197
Marx discussed (1997/2010), in which technological systems are understood as autonomous and
placed outside the realm of human relations.
At the same time, these narratives of naturalized disruption and replacement confront what I
have phrased in this chapter as ‘infrastructural expectations,’ or normative beliefs about how
infrastructure should function. These pieces show some of the infrastructural expectations that are
brought to platforms as digital memory infrastructure. The infrastructural expectations brought to
digital memory infrastructure are characterized by beliefs that infrastructure functions seamlessly and
invisibly, characteristics in line with wealthy nations’ infrastructural norms (Carse, 2016; Edwards,
2003), as well as more specific beliefs surrounding platforms as places of persistent content storage.
Press chronicling the breaking of digital memory infrastructure through sunsetting shows an
unresolved tension between ideologies of disruption and innovation and assumptions about
platforms’ relatively stable presences in daily life. Many of these accounts implicitly frame ‘what
happens to Facebook’ as what will eventually resolve these tensions, either proving out disruption
and replacement ideologies or reassuring infrastructural expectations of permanence and stability.
The breaking of digital memory infrastructure through sunsetting matters in these accounts–
perhaps not a surprise, since they are reporting on it. But who it is understood as mattering for–a sort
of cui bono–varies. First, there is an implicit understanding in press pieces discussing the economic
conditions of sunsetting that breakage is part of cycles of disruption, and content loss is a byproduct
of actions taken for the financial health of the company. For those who address the impact of
content loss explicitly from a ‘cultural’ value angle, sunsetting is phrased as mattering for the
individual, for families, and for the historical record, angles of importance I indeed have argued
throughout this dissertation. The solution to the issues platform sunsetting presents is typically
found in US archival organizations’ rescuing of social media platform data, content accrued with a
universalizing ‘capture it all’ ethos (De Kosnik, 2016). Importantly, there is little discussion in these

198
articles of the differential effects of content erasure and preservation for groups historically excluded
from historical self-representation, whether along racialized, geographic, or gendered lines, groups
who tend to be at the losing end of the ‘asymmetrical’ power relations that scholars have described
vis a vis data practice (D’Ignazio & Klein, 2020). Indeed, differential power relations are addressed
in just the one article (Krishna, 2019).
The universalizing perspective envisioned as a solution to the vagaries of content deletion
under sunsetting mimics early utopic narratives of the internet, which envisioned the internet as a
naturally egalitarian space where all of mankind would benefit from freedom of expression and
dissolved barriers to community (e.g. Barlow, 1996). These narratives were carried forward to social
media platforms, expounded upon by companies who framed themselves as providing a neutral
foundation for users to create and build global community (Gillespie, 2010). As critical technology
scholars have shown, though, universalizing narratives like these in fact privilege the invisible,
universal subject position–which is western, heterosexual, white, and male (De Kosnik, 2016; Tynes
& Noble, 2016). These narratives thus bring us full circle: back to the subject position that was the
imagined user of Vannevar Bush’s memex, that device that existed to perpetually represent men of
science’s trails of thought (1945). Discussions of ethical sunsetting and platform preservation,
whether public or academic, then, must acknowledge social power first, lest the future of the past be
limited to these visions of man, and memory, and machine.

199
Conclusion

hi
I have an internet question
So read a string of texts a friend sent me on an October Saturday in 2020. Like most friends I have
outside academia, she had a vague idea of what I did research on: the internet, even social media.
She didn’t know the extent to which I was focusing on the subject that she had chosen to text me
about:
is there any way to recover the mp3s that I uploaded to myspace music
circa 2007?
aka my band
I had interviewed five MySpace employees that week, and it was becoming clear to me that I would
never know the exact play-by-play of how its 2019 data loss had happened. I did know, however,
that there was little hope in recovering her band’s music. I summarized the bad news.
Those texts had come at an opportune time: faced with the nonchalant attitude toward
content loss I had been hearing in those interviews, I was left wondering whether my interviewees
were right in mapping their own ambivalent attitudes toward content loss onto their users. Maybe
no one did care. Maybe people were grateful when the shoebox was burned, as employee Gavin had
noted–and when someone made the decision to burn it for you.
Or maybe they weren’t:
TELL THEM THEY’VE ERASED MY BAND ARCHIVE

200
I didn’t relay her message to my future interviewees, but her exclamation had raised an interesting
question: was there even anyone to tell? Would it matter if I did? To put it another way: is there
accountability for platforms during sunsetting? Is there hope for responsible data stewardship even
in moments of decline, disrepair, and dismantling? To conclude this dissertation, I consider these
questions in the context of sunsetting as explored in the preceding five chapters, and in relation to
insights about memory, social media, and socio-technical aspects of dismantling that they presented.
This project has first attempted, as my friend asked, to tell: to make visible the process of
data erasure and preservation as social media platforms shut down. Describing the technical, value-
laden, and valuation-informed aspects of platform closure at the production-level, it has worked to
undo the ‘aura of phantom objectivity,’ as Leo Marx has phrased it (2010/1997), that surrounds
platforms in general and platform closure in particular.
As this project shows, platform sunsetting is not uniform. Instead, varying social and
technical conditions (and decisions) determine the path sunsetting takes and the magnitude of
commercial digital loss that is generated as it proceeds. MySpace’s content losses in 2013 and in
2019, for instance, were shaped by early technical choices, by perceptions of diminished and low-
value community as the platform aged, and by economic conditions around relaunching a new
platform that could extract saleable insights from user data. This resulted in a sunsetting process and
scale of loss that differed from Vine, whose employees drew on personal values around user
community, as well as economic reasoning around why deleting content would be bad for the larger
Twitter organization, to advocate for a sunsetting that did not result in the same magnitude of digital
loss.
This variability in sunsetting shown here, however, belies popular language about technical
closure that communicates its homogeneity. That is, while breakdown is rarely addressed in industry,
public, or academic discourse when compared to innovation, when it is, it is typically referred to as a

201
phase in a system’s lifecycle. An online business course teaching about technology and innovation
notes that “the technology life cycle has four distinct stages: research and development, ascent,
maturity, and decline” (Lumen, n.d.), a Google-made course on data management compares the
“data lifecycle” to the lifecycle of a butterfly (‘Stages of the data lifecycle,’ n.d.), and texts from
business and engineering discuss lifecycles as inevitable parts of building a data product (e.g.
Bloomberg Professional Services, 2015). Likewise, Chapter 5 described popular press’ framing of
social media platforms as having lifecycles (while also being subject to natural selection), just as I
have occasionally used this term as a convenient expression for describing a platform over time. The
idea that platforms and other technological systems have a lifecycle, however, presents decline and
closure, deletion and breakdown, as uniform, naturalized and subject to inevitable repeated phases––
that is, as ruled by ‘phantom objectivity.’ By describing sunsetting in detail at the production-level,
this project works to de-naturalize the ‘lifecycles’ of information systems, showing how these
processes are not identical nor inevitable. Instead, they are constructed through human relations,
relations that are mutable, as can be seen when comparing cases like Vine and MySpace. In revealing
these relations, the dissertation establishes a foundation for envisioning how a system’s ‘end of life’
could be constructed otherwise.
This project has also shown that platforms have sociotechnical entanglements, and social
effects, beyond the moments of innovation and active use that are typically the focus of public and
scholarly inquiry. They have these social effects because platforms are core components of digital
memory infrastructure, or the digital computational systems that underly, support, and shape
modern memory practice. Because personal and cultural memory practices are intimately tied to
media and those media’s preservation over time, patterns of loss or attempts at preservation during
sunsetting will come to bear on mediated reflections on the past.

202
As anthropologist Michel-Rolph Trouillot (1995) has argued, historical production encodes
silences that are informed by systems of power. Questions around which stories are forgotten, which
are told, who has the choice to speak about the past or to remain silent, are all deeply enmeshed in
racialized, gendered, and geopolitical forms of domination. As has been shown here, just as the
economic contexts and cultural ideologies of platform designers shape the politics of extant social
media platforms (Van Dijck et al, 2018), so too do their perspectives shape the uneven effects of
digital memory infrastructure in decline and disrepair. This project has shown that economic modes
of valuation endemic to US technology cultures in particular are crucial in creating these mediated
silences through commercial digital loss.
The bending of public discourse and social representation toward those who already have
power is not new, nor are the creation of silences (willful or not) as part of representational control
(e.g. Caswell, 2014; Medina, 2018). In that way, this project does not draw attention to a form of
representational power without precedent. What it does show, however, are dynamics through
which silences are created today, and the important role that social media platforms–as systems that
store and organize media and as systems that are expected to store and organize media–play in shaping
representation over time. It does show how social media platforms during and after sunsetting can
also embody, intensify, and accelerate existing power dynamics, all while making these dynamics
visible in different ways. The relative technical and social ease with which large-scale loss can be
enacted–deleting content produced largely by non-elites–and the opacity surrounding how loss is
created that are shown in this project are cases in point.
Toward Deletion Studies
In arguing that social media platforms shape public discourse over time by constructing
commercially inflected digital loss, this project has demonstrated that information deletion is not an
arbitrary nor pre-ordained byproduct of the ‘end of life’ stage of a platform. Instead, deletion is a

203
processual phenomenon intwined with dynamics of power, sociality, and materiality. Focusing on
media deletion (rather than creation) on social media platforms has in particular highlighted the
asymmetrical, commercially shaped informational control that platform organizations ultimately
wield over user-created media.
The utility of looking to deletion in this project points toward a broader research nexus that
I am referring to as deletion studies. Deletion studies broadly considers the ways in which processes of
removal, absence, erasure, and destruction modulate the social functions of information and
information technologies. In line with the analytical recasting that scholars like Steven Jackson have
suggested with ‘broken world thinking’ (2014), or the figure-ground shifts that Michel-Rolph
Trouillot (1995) argues for in looking toward the construction of historiographical ‘silence,’ a focus
on deletion may be a similarly useful inversion for deepening an understanding of sociotechnical
information systems today. This project has served as an example of what might come from studies
on deletion, as it has examined informational destruction through sunsetting in the context of social
media platforms. To conclude, however, I would like to briefly point toward where other studies of
deletion have proved useful in deepening understandings of information systems, especially
platforms; offer paths for where future studies of deletion may draw from and look to; and argue for
the usefulness of concatenating otherwise disparate studies of informational erasure under the
banner of deletion studies.
While deletion is not typically foregrounded as a common thread in platforms research, it is
nonetheless at the heart of some of the most heated public and academic debates about these
systems today. Content moderation on sites like Facebook and Twitter, for instance, has received
significant attention in scholarly and public press, literature that has revealed the ideologies
(Gillespie, 2018), labor force (Roberts, 2021), legal ambiguities (Gillespie, 2017b) and asymmetrical
agency (Myers West, 2018) surrounding the removal of content to administer (often harmful) speech

204
on these sites. Similarly, the act of ‘deplatforming’–or removing a person’s account from a platform–
has been highlighted following President Trump’s ban from Twitter in 2021 (Conger & Isaac, 2021),
bringing out contentious arguments over free speech and the social responsibility of platforms in its
wake. At the same time, deletion can also be employed by individual platform users as a form of
agency: users may delete previously posted content as an act of impression management, out of
privacy concerns, or to better reflect a changing identity (boyd, 2014; Christian, 2016). Data deletion
has been used too as part of larger scale political action, especially in the form of platform refusal
and non-use. In the United States, for instance, account deletion became minor political movement
under the hashtag #deleteFacebook, formed in reaction to the Cambridge Analytica scandal in which
tens of millions of Facebook profiles were mined for voter insights for the US 2016 election (Chen,
2018).
This is just a sample of the ways that deletion is wielded in multiple ways toward multiple
ends in relation to social media platforms. What bringing these disparate acts of information
destruction and erasure highlight for platform studies are the ways that platforms’ social power is
shaped significantly not just by what is uploaded and accessed on a particular site, but by what is
removed, erased, deleted. Focusing on deletion explicitly draws attention to who has the power to
shape these discourses and the mechanisms through which these discourses are shaped.
While platforms are the origin point for thinking about acts of information deletion, deletion
studies inevitably extend beyond these sites toward diverse contexts, whether the deaccessioning of
archival material (e.g. Radford, Radford & Lingel, 2012), the destruction of information and its
relationship to censorship (e.g. Liu & Zhao, 2021), the politics of training data deprecation in
machine learning contexts (e.g. Luccioni et al, 2022), and beyond. This examination of sunsetting
thus establishes a foundation for further research both within and outside of platform studies,

205
arguing that deletion is an especially salient node for highlighting sociotechnical informational
control.
Alternatives: Stewardship and Speculation
One goal in mapping how platforms arrive at deletion in this project was to provide a basis
for discussing what the long-term fate of digital cultural production hosted on platforms should look
like, offering an opening for considering alternatives beside what seemed to be the default mode of
content erasure under conditions of platform decline. The conclusions drawn from this project
present an opportunity to imagine how sunsetting could proceed otherwise, to think through what
long-term stewardship of platform data may look like, and to speculate about the potential value of
these traces outside of the dominant forms of economic valuation seen in this project.
Data stewardship is one way to envision these alternatives. This framing comes from Luke
Stark and Anna Lauren Hoffman’s (2019) analysis of codes of ethics in data science: noting that
while other professional codes–from fields like forestry and sanitation–use stewardship and
responsibility as driving values, data science models tend to forgo this framing, instead focusing on
‘top-down’ values like fairness or technical credibility. Just as Stark and Hoffman (2019) identify this
gap in data science, this project has likewise identified a dearth of industry-level conversation around
data stewardship when it comes to platforms. While I do not sketch out the multitude of concrete
forms stewardship could take here, it is clear that sunsetting may proceed otherwise if it was framed
through the lens of stewardship, a lens that resonates with Vine employees’ feelings of responsibility
toward Vine users and Vine content.
In particular, “collaborative stewardship,” a term I borrow from discussions of ethics in the
field of archaeology (Wylie, 2005), would be a logical starting place. Collaborative stewardship
involves the negotiated “co-management among divergent interests” of cultural heritage (Wylie,
2005, p. 65), a type of stewardship that could represent the economic needs of the platform

206
organization, yes, but also needs of users, the counsel of archivists, the perspective of historians, and
beyond. Like in archaeology, collaborative stewardship does not necessarily mean unequivocal
preservation (Harding, 1997; cited in Matthes, 2018): but it does mean that platform organizations
would pause to consider what and whose values guide their decisions during closure.
One way that conversations re-imagining sunsetting could begin is by reframing the
processes of valuation that are applied to content at a platform’s end of life. While organizations
speculated about what financial value could be extracted from varying components when a platform
was in decline–asking what components could be worth–speculation as a critical process can be used
to imagine alternatives (Sheppard, 2017), in this case, alternatives to the default of commercial digital
loss.
28
The archiving of GeoCities by the Internet Archive offers a hopeful starting point. When
GeoCities announced its closure in 2009, the Internet Archive was able to capture significant
portions of its pages, with the public nominating specific pages to be archived. This collection has
led toward diverse uses, including academic inquiry (Milligan, 2015), art (Espenschied & Lialina,
2010), or simply reminiscence (‘blade’s place,’ n.d.). The archival organization Documenting the
Now offers another path of generative speculation: the group focuses on community-centered
approaches toward archiving the web and social media, gathering this data as a historical resource as
well as for use by community and activist groups in memory work (‘Documenting the Now
Project’).
This is ultimately an optimistic vision of collaborative data stewardship, and there are
realistic constraints when considering this approach. There are indeed differences between the
archaeological context in which collaborative stewardship was developed and the platforms context

28
As the cinema scholar Samantha Sheppard (2017) has noted, speculative practice–like creating fictional
archives and imagined classic films from the history of Black American cinema–has been a crucial
methodology for reshaping a past whose representative works were lost, marginalized, or forgotten.

207
to which it is applied here. Archaeology is fundamentally preservation-oriented, with profit rarely a
driving force for the field within the context of not-for-profit museums and universities. Platforms,
on the other hand, are more readily aligned with corporate profit-seeking; as Robert Gorwa (2019, p.
860) has put it in his discussion of platform governance, “platform companies are companies.”
Moreover, even when a not-for-profit, stewardship-focused organization is called into the sunsetting
process, it does not always lead to a robust and perpetual archival record. In the case of Vine, for
instance, the Internet Archive required a million-dollar donation for the organization to take on
Vine’s content in perpetuity. An interview with Jefferson Bailey, the Director of Web Archiving at
the Internet Archive, further revealed the financial constraints that shape preservation. Collecting
social media data, he noted, does not mean that data will be readily publicly accessible; making these
archives searchable and user-friendly is only possible when there are financial and temporal
resources at the archive’s disposal. This is to say that thinking about sunsetting in terms of
collaborative stewardship requires the acknowledgement that the “co-management of divergent
interests” (Wylie, 2005) includes the commercial-corporate interests of platform companies and
venture capitalists, the financial and temporal constraints of archival organizations, as well as the
priorities of users (and user communities), platform employees, artists, academics, historians,
governments, and beyond.
The cases examined in this project show one way that commercial priorities can be aligned
with responsible data preservation. In Vine’s case, employees lobbied for preservation by arguing
that content preservation made good business sense for Twitter, who did not want to suffer more
backlash than it already had for deleting Vine. Indeed, most of these sites are owned by companies
that survive past the individual platform’s shutdown, whether Twitter’s existence after Vine, Yahoo’s
survival past GeoCities, or Specific/Viant’s endurance past MySpace. Continued ownership provides
a foundation for these companies to be invested in preservation as part of ongoing reputation

208
management. Moreover, information preservation for reputation management is already a known
practice for large, longstanding corporations. Corporate historians are often hired to help employees
understand a brand’s evolution, document the company over time, and help cement the
organization’s legacy (Koshy, 2017). Sometimes their roles include contextualizing business failures
as part of this overarching story (e.g. “Chronological History of IBM: 1990s”). At the same time, this
reputation-based sense of responsibility requires broader public awareness of sunsetting such that
companies believe stewardship during the shutdown process matters to their public standing.
To this end, the ongoing and prominent public, governmental, and scholarly conversations
around platform governance may do well to incorporate discussions of platform sunsetting.
Addressing the rights and responsibilities of these organizations has led to both formal legal
frameworks (e.g. GDPR) as well as the development of corporate practices (e.g. greater transparency
around content moderation in user interfaces), tactics that were shaped by an evolving cultural and
scholarly understanding of the public roles of communication platforms. Incorporating sunsetting
into platform governance conversations would not only serve to deepen these discussions, reflecting
the ways that platforms continue to have social effects even after decline, but also would lay the
foundation for formalizing responsible sunsetting practices across the platform industry. Like other
platform governance approaches, these tactics may be legal in nature–for instance, adding sunsetting
stipulations to emerging data governance laws, like a 90-day window between a sunsetting
announcement and content deletion–just as they may be based in corporate best practices, like
planning for preservation costs when projecting a company’s financial future. Already incorporating
the perspectives of multiple stakeholders (Gorwa, 2019), platform governance conversations serve
as a generative complement to notions of collaborative stewardship during platform sunsetting.
Sunsetting ostensibly means an end. When viewed through the speculative possibilities of
stewardship, however, sunsetting also hints at persistence and even new beginnings. Should content

209
be preserved, new practices, new uses, and new forms of value may emerge from what came before.
In the end, that is, the sun may also rise.

210
References
Adorno, T., & Horkheimer, M. (2002). Dialectic of Enlightenment. Stanford University Press.
Ahmed, S. I., Jackson, S. J., & Rifat, Md. R. (2015). Learning to fix: Knowledge, collaboration and
mobile phone repair in Dhaka, Bangladesh. Proceedings of the Seventh International Conference on
Information and Communication Technologies and Development, 1–10.
https://doi.org/10.1145/2737856.2738018
Alarcon, A., Altrudi, S., Corry, F., & Forelle, M. (Forthcoming). The Sociotechnical Imaginaries of
1968. International Journal of Communication.
Alexander, C., & Sanchez, A. (2018). Indeterminacy: Waste, Value, and the Imagination (1st ed., Vol. 7).
Berghahn Books. https://doi.org/10.2307/j.ctvw04gbz
Allington-Smith, D. (2015). Anticipating the Internet: How the predictions of Paul Otlet, H.G. Wells and
Vannevar Bush shaped the Digital Information Age. City University London.
Anand, N., Gupta, A., & Appel, H. (Eds.). (2018). The Promise of Infrastructure. Duke University Press.
https://doi.org/10.1215/9781478002031
Andor Brodeur, M. (2016, October 28). Requiem for Vine. Boston Globe, G4.
Andreessen, M. (2007, October 18). The three kinds of platforms you meet on the Internet.
https://web.archive.org/web/20071018161644/http://blog.pmarca.com/2007/09/the-three-
kinds.html
Andrejevic, M. (2009). Exploiting YouTube: Contradictions of user-generated labour. In P. Snickers
& P. Vonderau (Eds.), The YouTube Reader. National Library of Sweden.
Ankerson, M. S. (2018). Dot-Com Design. NYU Press.
Appadurai, A. (1986). The Social life of things: Commodities in cultural perspective. Cambridge University
Press.

211
Arango, T. (2011, January 11). After Failing to Innovate, MySpace Struggles. The New York Times.
https://www.nytimes.com/2011/01/12/technology/internet/12myspace.html
Arrington, M. (2006a). Friendster Recapitalized. TechCrunch.
https://social.techcrunch.com/2006/02/02/friendster-recapitalized/
Arrington, M. (2006b, December 12). It’s Official(ish): MySpace Is Biggest Site on Internet.
TechCrunch. https://social.techcrunch.com/2006/12/12/its-officialish-myspace-is-biggest-site-
on-internet/
Assmann, A. (1996). Texts, Traces, Trash: The Changing Media of Cultural Memory. Representations,
56, 123–134. https://doi.org/10.2307/2928711
Assmann, J. (2008). Communicative and Cultural Memory. In A. Erll & A. Nünning (Eds.), Cultural
Memory Studies: An International and Interdisciplinary Handbook (pp. 109–118). De Gruyter.
Assmann, J., & Czaplicka, J. (1995). Collective Memory and Cultural Identity. New German Critique,
65, 125–133. JSTOR. https://doi.org/10.2307/488538
Atkinson, P. (2014). ENIAC versus Colossus and the early presentation of electronic computers.
Unpublished. 2014 Annual Design History Society Conference, University of Oxford.
http://shura.shu.ac.uk/9501/3/Atkinson_-_Eniac_versus_Colossus_paper_%26_url.pdf
Avance, R. (2016). Community. In B. Peters (Ed.), Digital Keywords (pp. 63–69). Princeton University
Press; JSTOR. https://doi.org/10.2307/j.ctvct0023.10
Axe, D. (2018). 15 Years Ago, the Military Tried to Record Whole Human Lives. It Ended Badly.
Vice. https://www.vice.com/en/article/vbqdb8/15-years-ago-the-military-tried-to-record-
whole-human-lives-it-ended-badly
Ballatore, A., & Natale, S. (2016). E-readers and the death of the book: Or, new media and the myth
of the disappearing medium. New Media & Society, 18(10), 2379–2394.
https://doi.org/10.1177/1461444815586984

212
Barlow, J. P. (2016, January 20). A Declaration of the Independence of Cyberspace. Electronic Frontier
Foundation. https://www.eff.org/cyberspace-independence
Barocas, S., Crawford, K., Shapiro, A., & Wallach, H. (2017). The Problem With Bias: Allocative
Versus Representational Harms in Machine Learning. SIGCIS Conference. SIGCIS.
http://meetings.sigcis.org/uploads/6/3/6/8/6368912/program.pdf
Beer, D. (2018). Envisioning the power of data analytics. Information, Communication & Society, 21(3),
465–479. https://doi.org/10.1080/1369118X.2017.1289232
Bell, G. (2001). A Personal Digital Store. Communications of the ACM, 44(1), 86–91.
Berger, P. L., & Luckmann, T. (1966). The social construction of reality: A treatise in the sociology of knowledge.
Penguin Books.
Blanchette, J.-F. (2011). A material history of bits. Journal of the American Society for Information Science
and Technology, 62(6), 1042–1057. https://doi.org/10.1002/asi.21542
Bloomberg Professional Services. (2015, July 14). 7 phases of a data life cycle. Bloomberg Professional
Services. https://www.bloomberg.com/professional/blog/7-phases-of-a-data-life-cycle/
Bolzoni, L. (2021). Art of Memory. In A. Blair, P. Duguid, A.-S. Goeing, & A. Grafton (Eds.),
Information: A Historical Companion (pp. 318–322). Princeton University Press.
Bostrom, N. (2003). Transhumanist Values. In F. Adams (Ed.), Ethical Issues for the 21st Century.
Philosophical Documentation Center Press.
Bourdieu, P., & Wacquant, L. (2013). Symbolic capital and social classes. Journal of Classical Sociology,
13(2), 292–302. https://doi.org/10.1177/1468795X12468736
Bourne, A. T., Caroline George, Joseph Kane, and Andrew. (2021, November 9). America has an
infrastructure bill. What happens next? Brookings. https://www.brookings.edu/blog/the-
avenue/2021/11/09/america-has-an-infrastructure-bill-what-happens-next/

213
Bowen, G. A. (2009). Document Analysis as a Qualitative Research Method. Qualitative Research
Journal, 9(2), 27–40. https://doi.org/10.3316/QRJ0902027
Bowker, G. C. (2006). Memory practices in the sciences. MIT Press.
boyd, danah. (2012). White Flight in Networked Publics: How Race and Class Shaped American
Teen Engagement with MySpace and Facebook. In L. Nakamura & P. Chow-White (Eds.),
Race After the Internet (pp. 209–228). Routledge.
boyd, danah. (2014). It’s complicated: The social lives of networked teens. Yale University Press.
Boym, S. (2001). The Future of Nostalgia. Basic Books.
Braun, V., & Clarke, V. (2006). Using thematic analysis in psychology. Qualitative Research in Psychology,
3(2), 77–101. https://doi.org/10.1191/1478088706qp063oa
Brex. (2021, July 21). How to Do a Startup Valuation Using 8 Different Methods.
https://www.brex.com/blog/startup-valuation/
Brown, E. (2017, March 3). Snapchat Investors Found Their Golden Goose in a Field of Turkeys;
Question hanging over Snap Inc.: Will it sputter like Twitter, or soar to new heights as next
media giant? The Wall Street Journal.
Brown, J. K. (2014). Not the Eads Bridge: An Exploration of Counterfactual History of Technology.
Technology and Culture, 55(3), 521–559.
Brunton, F. (2013). Spam: A shadow history of the Internet. MIT Press.
Brunton, F., & Coleman, G. (2014). Closer to the Metal. In K. A. Foot, P. J. Boczkowski, & T.
Gillespie (Eds.), Media Technologies: Essays on Communication, Materiality, and Society (pp. 77–97).
MIT Press.
Bush, V. (1945). As We May Think. The Atlantic Monthly, 176(1), 101–108.

214
Calhoun, K. (2019). Vine Racial Comedy as Anti-Hegemonic Humor: Linguistic Performance and
Generic Innovation. Journal of Linguistic Anthropology, 29(1), 27–49.
https://doi.org/10.1111/jola.12206
Callon, M. (1984). Some elements of a sociology of translation: Domestication of the scallops and
the fishermen of St Brieuc Bay. The Sociological Review, 32(1_suppl), 196–233.
Carse, A. (2016). Keyword: Infrastructure – How a Humble French Engineering Term Shaped the
Modern World. In P. Harvey, C. B. Jensen, & A. Morita (Eds.), Infrastructures and Social
Complexity: A Routledge Companion. Routledge.
Cashmore, E. (2002). The Black Culture Industry. Routledge. https://doi.org/10.4324/9780203132043
Castells, M. (2010). The Rise of the Network Society (Vol. 1). John Wiley & Son.
Caswell, M. (2014). Archiving the Unspeakable: Silence, Memory, and the Photographic Record in Cambodia.
University of Wisconsin Pres.
Chokshi, N. (2019, March 19). MySpace, Once the King of Social Networks, Lost Years of Data
From Its Heyday. New York Times. https://www.nytimes.com/2019/03/19/business/myspace-
user-data.html
Christin, A. (2020). Metrics at Work: Journalism and the Contested Meaning of Algorithms. Princeton
University Press.
Chun, W. (2011). Programmed Visions. MIT Press. https://mitpress.mit.edu/books/programmed-
visions
Chun, W. H. K. (2011). Programmed Visions: Software and Memory. MIT Press.
Chung, A. (2011, March 8). Wal-Mart workers hope to show their union the door. The Toronto Star.
https://www.thestar.com/news/canada/2011/03/08/walmart_workers_hope_to_show_their
_union_the_door.html

215
Clark, J., Couldry, N., De Kosnik, A., Gillespie, T., Jenkins, H., Kelty, C., Papacharissi, Z., Powell,
A., & Van Dijck, J. (2014). Part 5: Platforms. International Journal of Communication, 8.
CNN. (2011, September 22). Facebook unveils “Timeline” feature.
https://www.youtube.com/watch?v=bRNQglXNIXc
Cohn, M. L. (2019). Keeping Software Present: Software as a Timely Object for STS Studies of the
Digital. In J. Vertesi, D. Ribes, C. DiSalvo, Y. Loukissas, L. Forlano, D. K. Rosner, S. J.
Jackson, & H. R. Shell (Eds.), DigitalSTS (pp. 423–446). Princeton University Press; JSTOR.
https://doi.org/10.2307/j.ctvc77mp9.31
Compton, N. (2020, May 4). Go ahead and post your old travel photos. Experts say it can improve
your mood. Washington Post. https://www.washingtonpost.com/travel/tips/go-ahead-post-
your-old-travel-photos-experts-say-it-can-improve-your-mood/
Computer History Museum. (n.d.). Magnetic Tape—CHM Revolution. Retrieved January 18, 2022, from
https://www.computerhistory.org/revolution/memory-storage/8/258
Construction. (n.d.). In Oxford English Dictionary.
Corbin, J., & Strauss, A. (1990). Grounded Theory Research: Procedures, Canons and Evaluative
Criteria. Zeitschrift Für Soziologie, 19(6), 418–427. https://doi.org/10.1515/zfsoz-1990-0602
Corry, F. (2020). ‘LADY U SEND ME YR MOVIE:’ Constructing Joanie 4 Jackie’s Feminist
Distribution Network. Feminist Media Studies. https://doi.org/10.1080/14680777.2020.1785912
Corry, F. (2021). Why does a platform die? Diagnosing platform death at Friendster’s end. Internet
Histories. https://doi.org/10.1080/24701475.2021.1985360
Cotter, C. (2015). Discourse and Media. 795–821. https://doi.org/10.1002/9781118584194.ch37
Couldry, N., & Powell, A. (2014). Big Data from the bottom up. Big Data & Society, 1(2),
2053951714539277. https://doi.org/10.1177/2053951714539277

216
Cowan, R. S. (1985). How the refrigerator got its hum. In D. A. MacKenzie & Judy. Wajcman
(Eds.), The Social Shaping of Technology. Open University Press.
Crang, M., Gregson, N., Ahamed, F., Ferdous, R., & Akhter, N. (2012). Death, the Phoenix and
Pandora: Transforming things and values in Bangladesh. In J. Reno & C. Alexander (Eds.),
Economies of recycling: The global transformations of materials, values and social relations (pp. 59–75). Zed
Books.
Crawford, S. P. (n.d.). InternetThink. Journal on Telecommunications and High Technology Law, 5, 467–486.
Crehan, K. A. F. (2016). Gramsci’s common sense: Inequality and its narratives. Duke University Press.
https://doi.org/10.1515/9780822373742
Chronological History of IBM: 1990s. (n.d.). IBM: History.
https://www.ibm.com/ibm/history/history/decade_1990.html
Crovitz, L. G. (2012, July 22). Gordon Crovitz: Who Really Invented the Internet? Wall Street Journal.
https://online.wsj.com/article/SB10000872396390444464304577539063008406518.html
Dalal, V., Krish, K., Munstermann, B., & Patenge, R. (2020, October 6). Tech debt: Reclaiming tech
equity. McKinsey. https://www.mckinsey.com/business-functions/mckinsey-digital/our-
insights/tech-debt-reclaiming-tech-equity
DARPA. (2020, May 7). Lifelog and Facebook [Tweet]. @DARPA.
https://twitter.com/DARPA/status/1258210075849756673
DARPA Information Processing Technology Office. (n.d.). [Archived] Lifelog. Retrieved January 18,
2022, from
https://web.archive.org/web/20040202205208/http:/www.darpa.mil/ipto/Programs/lifelog/i
ndex.htm

217
Daubs, M. S. (2020). User-generated content. In D. Merskin (Ed.), The SAGE international encyclopedia
of mass media and society (Vol. 1, pp. 1826–1827). SAGE Publications.
https://dx.doi.org/10.4135/9781483375519.n705
De Kosnik, A. (2016). Rogue Archives: Digital Cultural Memory and Media Fandom. MIT Press.
de Vise, D. (2011, July 18). A bit of time traveling in the Internet age. The Washington Post, B1.
Deathwatch—Archiveteam. (n.d.). Retrieved April 24, 2022, from
https://wiki.archiveteam.org/index.php/Deathwatch
Derrida, J. (1995). Archive Fever: A Freudian Impression (E. Prenowitz, Trans.). Diacritics, 25(2), 9–
63.
Deseriis, M. (2015). Improper Names: Collective Pseudonyms from the Luddites to Anonymous. University of
Minnesota Press. https://doi.org/10.5749/j.ctt173zm71
Deutsch, W. (2017, February 27). The Elusive Hockey-Stick Sales Curve. The University of Chicago
Booth School of Business. https://www.chicagobooth.edu/review/elusive-hockey-stick-sales-curve
D’Ignazio, C., & Klein, L. F. (2020). Data Feminism. MIT Press.
https://doi.org/10.7551/mitpress/11805.001.0001
Dilley, J., Maggs, B., Parikh, J., Prokop, H., Sitaraman, R., & Weihl, B. (2002). Globally distributed
content delivery. IEEE Internet Computing, 6(5), 50–58.
https://doi.org/10.1109/MIC.2002.1036038
Dix, A. (2002). The Ultimate Interface and the Sums of Life? Interfaces, 50, 16.
Documenting the Now. (n.d.). DocNow. Retrieved April 21, 2022, from https://www.docnow.io/
Dow Up; Bailout for Portugal; Facebook Too Big to Fail? (2010, December 1). Finance Wire.
http://libproxy.usc.edu/login?url=https://www.proquest.com/wire-feeds/dow-upbailout-
portugal-facebook-too-big-fail/docview/815374686/se-2?accountid=14749

218
Draper, A. F. (1953). Univac on election night. Electrical Engineering, 72(4), 291–293.
https://doi.org/10.1109/EE.1953.6438592
Draper, N. (2017). Fail Fast: The Value of Studying Unsuccessful Technology Companies[1]. Media
Industries Journal, 4(1). https://doi.org/10.3998/mij.15031809.0004.101
Drozdzewski, D., & Birdsall, C. (2019). Advancing Memory Methods. In D. Drozdzewski & C.
Birdsall (Eds.), Doing Memory Research: New Methods and Approaches (pp. 1–20). Springer
Singapore. https://doi.org/10.1007/978-981-13-1411-7_1
Dumenco, S. (2012, May 14). It’s curtains for Google! And Facebook! And Tumblr! And ...
Advertising Age, 83(20), 31.
Dunbar–Hester, C. (2014). Low Power to the People: Pirates, Protest, and Politics in FM Radio Activism. MIT
Press.
Dyer, R. (2008). The Matter of Whiteness. In P. S. Rothenberg (Ed.), White privilege: Essential readings
on the other side of racism (3rd ed., pp. 9–14). Worth Publishers.
Edwards, P. N. (2003). Infrastructure and Modernity: Force, Time, and Social Organization in the
History of Sociotechnical Systems. In T. J. Misa, A. Feenberg, & P. Brey (Eds.), Modernity and
Technology. MIT Press.
Edwards, P. N. (2019). Infrastructuration: On Habits, Norms and Routines as Elements of
Infrastructure. Thinking Infrastructures, 62, 355–366. https://doi.org/10.1108/S0733-
558X20190000062022
Edwards, P. N., Jackson, S. J., Chalmers, M. K., Bowker, G. C., Borgman, C. L., Ribes, D., Burton,
M., & Calvert, S. (2012). Knowledge Infrastructures: Intellectual Frameworks and Research Challenges.
http://pne.people.si.umich.edu/PDF/Edwards_etal_2013_Knowledge_Infrastructures.pdf
El Akkad, O. (2013, November 22). A requiem for Winamp, the best MP3-player ever made. The
Globe and Mail.

219
Electronic Brain Records Causes of Plane Crashes. (1944, November 15). St. Louis Post - Dispatch.
Elgan, M. (2016, April 4). Lifelogging is dead (for now). Computerworld.
https://www.computerworld.com/article/3048497/lifelogging-is-dead-for-now.html
Ensmenger, N. (2014, September 11). When Good Software Goes Bad. Nathan Ensmenger Website.
https://homes.luddy.indiana.edu/nensmeng/files/ensmenger-mice.pdf
Eriksen, T. H., & Schober, E. (2017). Waste and the superfluous: An introduction. Social
Anthropology, 25(3), 282–287. https://doi.org/10.1111/1469-8676.12422
Erin, A. (2016, October 29). If Twitter doesn’t tidy up, it will go the way of Vine. The Globe and Mail,
B8.
Fisher, D., & Smiley, B. (2016). Adapting Rubbish Theory for heritage tourism. Journal of Heritage
Tourism, 11(2), 143–154. https://doi.org/10.1080/1743873X.2015.1080711
Foot, K. A., Boczkowski, P. J., & Gillespie, T. (Eds.). (2014). Media Technologies: Essays on
Communication, Materiality, and Society. MIT Press.
Friedman, B., Kahn, P. H., & Borning, A. (2006). Value Sensitive Design and Information Systems.
In P. Zhang & D. F. Galletta (Eds.), Human-computer interaction and management information systems:
Foundations (pp. 348–372). M.E. Sharpe.
Friend, T. (2014, December 8). Hollywood and Vine. The New Yorker.
http://www.newyorker.com/magazine/2014/12/15/hollywood-vine
Gabrys, J. (2011). Digital Rubbish: A Natural History of Electronics (p. ix). University of Michigan Press.
https://doi.org/10.3998/dcbooks.9380304.0001.001
Gandy, O. (1996). Coming to Terms with the Panoptic Sort. In D. Lyon & E. Zureik (Eds.),
Computers, surveillance, and privacy. University of Minnesota Press.
Gannes, L. (2010, August 4). Facebook Buys Friendster Patents for $40M. Gigaom.
https://gigaom.com/2010/08/04/facebook-buys-friendster-patents-for-40m/

220
Geiger, R. S. (2014). Bots, bespoke, code and the materiality of software platforms. Information,
Communication & Society, 17(3), 342–356. https://doi.org/10.1080/1369118X.2013.873069
Gemmell, J., Bell, G., & Lueder, R. (n.d.). MyLifeBits project. Retrieved January 18, 2022, from
https://slideplayer.com/slide/8582715/
Gemmell, J., Bell, G., & Lueder, R. (2006). MyLifeBits: A personal database for everything.
Communications of the ACM, 49(1), 88–95.
GeoHoods @ Blade’s Place—GeoCities Homestead Neighborhoods and Suburbs. (n.d.). Retrieved April 21,
2022, from https://www.bladesplace.id.au/geocities-neighborhoods-suburbs.html
Gillespie, T. (2010). The politics of ‘platforms.’ New Media & Society, 12(3), 347–364.
https://doi.org/10.1177/1461444809342738
Gillespie, T. (2017a). Governance of and by Platforms. In J. Burgess, T. Poell, & A. Marwick (Eds.),
SAGE Handbook of Social Media. SAGE.
Gillespie, T. (2017b). Platforms are not intermediaries. Geo. L. Tech. Rev., 2, 198.
Gillespie, T. (2018). Custodians of the Internet: Platforms, content moderation and the hidden decisions that shape
social media. Yale University Press.
Gillette, F. (2011, June 22). The Rise and Inglorious Fall of MySpace. Bloomberg.
https://www.bloomberg.com/news/articles/2011-06-22/the-rise-and-inglorious-fall-of-
myspace
Gipson, B., Corry, F., & Noble, S. (2021). Intersectionality (pp. 305–312).
https://doi.org/10.7551/mitpress/12236.003.0033
Glaser, B. G., & Strauss, A. L. (1967). The discovery of grounded theory; strategies for qualitative research.
Aldine Pub. Co.

221
Goodin, E. (2021, December 7). Biden’s picks for FCC and and FTC are slammed for old tweets. Mail
Online. https://www.dailymail.co.uk/news/article-10281783/Bidens-picks-FCC-FTC-accused-
political-old-tweets-resurface.html
Google. (n.d.). Stages of the data life cycle. Coursera. Retrieved April 24, 2022, from
https://www.coursera.org/lecture/foundations-data/stages-of-the-data-life-cycle-mhEHY
Gotthart, C. (1944, December 9). Experts Reveal Marvels of B-29 Gunnery Brain. Chicago Daily
Tribune.
Graham, S., & Marvin, S. (2001). Splintering urbanism: Networked infrastructures, technological mobilites and
the urban condition. Routledge.
Gramsci, A. (2011). Prison notebooks (J. A. Buttigieg & Antonio. Callari, Eds.; 2011 pbk. ed.).
Columbia University Press.
Greene, D., & Shilton, K. (2018). Platform privacies: Governance, collaboration, and the different
meanings of “privacy” in iOS and Android development. New Media & Society, 20(4), 1640–
1657. https://doi.org/10.1177/1461444817702397
Greeson, E. (2020). Ecologies of Valuation. Valuation Studies, 7(2).
https://doi.org/10.3384/VS.2001-5992.2020.7.2.167-196
Greeson, E., Laser, S., & Pyyhtinen, O. (2020). Dis/Assembling Value: Lessons from Waste
Valuation Practices. Valuation Studies, 7(2). https://doi.org/10.3384/VS.2001-
5992.2020.7.2.151-166
Hagiu, A., & Rothman, S. (2016, April 1). Network Effects Aren’t Enough. Harvard Business Review.
https://hbr.org/2016/04/network-effects-arent-enough
Halbwachs, M. (1992). On Collective Memory (L. Coser, Trans.). The University of Chicago Press.

222
Hansell, S. (2001, December 17). Technology & Media; For Some Dot-Coms, There Are Real
Profits. The New York Times. https://www.nytimes.com/2001/12/17/business/technology-
media-for-some-dot-coms-there-are-real-profits.html
Hansell, S. (2006, April 23). For MySpace, Making Friends Was Easy. Big Profit Is Tougher. The New
York Times. https://www.nytimes.com/2006/04/23/business/yourmoney/23myspace.html
Harrison, S. (n.d.). Five Years of Tech Diversity Reports—And Little Progress. Wired. Retrieved
April 11, 2022, from https://www.wired.com/story/five-years-tech-diversity-reports-little-
progress/
Hartman, S. (2007). Lose your mother: A journey along the Atlantic slave route. Farrar, Straus and Giroux.
Hawkins, G., & Muecke, S. (2003). Culture and waste: The creation and destruction of value. The Rowman
& Littlefield Publishing Group.
Heilemann, J. (2004, March 1). What’s Friendster Selling? Fortune.
https://archive.fortune.com/magazines/business2/business2_archive/2004/03/01/363554/in
dex.htm
Helmond, A., Nieborg, D. B., & ven der Vlist, F. N. (2019). Facebook’s evolution: Development of
a platform-as-infrastructure. Internet Histories, 3(2), 123–146.
Helmond, A., & Vlist, F. van der. (2019). Social Media and Platform Historiography: Challenges and
Opportunities. TMG Journal for Media History, 22(1), 6–34. https://doi.org/10.18146/tmg.434
Herzog, W. (2003). Wheel of Time. Werner Herzog Filmproduktion.
Hesmondhalgh, D. (2010). User-Generated Content, Free Labour and the Cultural Industries.
Ephemera: Theory & Politics in Organization, 19(3/4), 267–284.
Highfield, T. (2016). Social media and everyday politics. Polity.
Hirsch, M. (2008). The Generation of Postmemory. Poetics Today, 29(1), 103–128.

223
Hogan, B., & Quan-Haase, A. (2010). Persistence and Change in Social Media. Bulletin of Science,
Technology & Society, 30(5), 309–315. https://doi.org/10.1177/0270467610380012
Hogan, M. (2015). Data flows and water woes: The Utah Data Center. Big Data & Society, 2(2),
205395171559242. https://doi.org/10.1177/2053951715592429
Houston, L., Jackson, S. J., Rosner, D. K., Ahmed, S. I., Young, M., & Kang, L. (2016). Values in
Repair. Proceedings of the 2016 CHI Conference on Human Factors in Computing Systems, 1403–1414.
https://doi.org/10.1145/2858036.2858470
How we forgot about the world’s most popular music website. (2019, March 20). The Daily Telegraph,
24.
Hughes, J. (2016, October 31). Vine Dries Up. Black Humor Loses a Home. The New York Times.
https://www.nytimes.com/2016/11/01/arts/vine-jay-versace-black-culture.html
Humphreys, L. (2020). Birthdays, anniversaries, and temporalities: Or how the past is represented as
relevant through on-this-date media. New Media & Society, 22(9), 1663–1679.
https://doi.org/10.1177/1461444820914874
Jackson, S. J. (2014). Rethinking Repair. In T. Gillespie, P. J. Boczkowski, & K. A. Foot (Eds.),
Media Technologies: Essays on Communication, Materiality, and Society. MIT Press.
Jackson, S. J. (2016). Speed, Time, Infrastructure.
https://doi.org/10.1093/acprof:oso/9780198782858.003.0012
Jacobsen, B. N., & Beer, D. (2021). Quantified Nostalgia: Social Media, Metrics, and Memory. Social
Media + Society, 7(2), 20563051211008824. https://doi.org/10.1177/20563051211008822
Jafarinaimi, N., Nathan, L., & Hargraves, I. (2015). Values as Hypotheses: Design, Inquiry, and the
Service of Values. Design Issues, 31(4), 91–104. https://doi.org/10.1162/DESI_a_00354
Jamison, M. (2018, March 28). Five things Facebook needs to know about platform economics.
American Enterprise Institute for Public Policy Research.

224
Jasanoff, S., & Kim, S.-H. (2009). Containing the Atom: Sociotechnical Imaginaries and Nuclear
Power in the United States and South Korea. Minerva, 47(2), 119–146. JSTOR.
Jason Edward Black. (2002). The “Mascotting” of Native America: Construction, Commodity, and
Assimilation. American Indian Quarterly, 26(4), 605–622. JSTOR.
Kang, C. (n.d.). Memories, hanging by Web’s delicate thread. The Washington Post, 30.
Kaun, A., & Stiernstedt, F. (2014). Facebook time: Technological and institutional affordances for
media memories. New Media & Society, 16(7), 1154–1168.
https://doi.org/10.1177/1461444814544001
Kavadlo, J. (2013). 9/11 Did Not Take Place: Apocalypse and Amnesia in Film and The Road. The
Popular Culture Studies Journal, 61.
Kelkar, S. (2017). Engineering a platform: The construction of interfaces, users, organizational roles,
and the division of labor. New Media & Society, 20(7), 2629–2646.
Kelty, C., & Erickson, S. (2015). The Durability of Software. In I. Kaldrack & M. Leeker (Eds.),
There is no Software, there are just Services. meson press.
Kelty, C. M. (2008). Two bits: The cultural significance of free software. Duke University Press.
https://doi.org/10.1515/9780822389002
Kirby, J. (2013, February 12). When Advertising Meets the Meme. Harvard Business Review.
https://hbr.org/2013/02/when-advertising-meets-the-mem
Kline, R., & Pinch, T. (1996). Users as Agents of Technological Change: The Social Construction of
the Automobile in the United States. Technology and Culture, 37(4), 763–795.
Kneese, T. (2019). Networked heirlooms: The affective and financial logics of digital estate planning.
Cultural Studies, 33(2), 297–324. https://doi.org/10.1080/09502386.2018.1466904

225
Koshy, T. (2017, November 29). Inside my job: The corporate historian. Friday Magazine.
https://fridaymagazine.ae/life-culture/people-profiles/inside-my-job-the-corporate-historian-
1.2131737
Kosnik, A. D. (2016). Rogue Archives: Digital Cultural Memory and Media Fandom. MIT Press.
Kraus, A. (n.d.). Handling 1.5 Billion Page Views Per Day Using ASP.NET 2.0. Andreas Kraus Blog.
Retrieved April 4, 2022, from https://www.andreas-kraus.net/blog/handling-15-billion-page-
views-per-day-using-aspnet-20/
Krishna, R. (2019, March 30). We can’t just delete websites – they’re a cultural record. The
Independent, 35.
Krukowski, D. (2019, March 20). History Disappeared When MySpace Lost 12 Years of Music, and
It Will Happen Again. Pitchfork. https://pitchfork.com/thepitch/history-disappeared-when-
myspace-lost-12-years-of-music-and-it-will-happen-again/
Kuny, T. (1997). A Digital Dark Ages? Challenges in the Preservation of Electronic Information.
63RD IFLA Council and General Conference, 12.
Landsberg, A. (2004). Introduction: Memory Modernity Mass Culture. In Prosthetic Memory: The
Transformation of American Remembrance in the Age of Mass Culture (pp. 1–24). Columbia University
Press.
Lapinski, T. (2006, June). Spammers and ex-cons, not Tom, made MySpace, says journalism student.
Valleywag. http://www.valleywag.com/tech/myspace/spammers-and-excons-not-tom-made-
myspace-says-journalism-student-170413.php
Larkin, B. (2013). The Politics and Poetics of Infrastructure. Annual Review of Anthropology, 42(1),
327–343. https://doi.org/10.1146/annurev-anthro-092412-155522
Lee, L. A. (2017). Black Twitter: A Response to Bias in Mainstream Media. Social Sciences, 6(1).
https://doi.org/10.3390/socsci6010026

226
Legal Information Institute. (2021, November). Sunset law. LII / Legal Information Institute.
https://www.law.cornell.edu/wex/sunset_law
Lepore, J. (2014, June 16). The Disruption Machine: What the gospel of innovation gets wrong. The
New Yorker.
Leslie, C. (2020). As We Could Have Thought: Deploying Historical Narratives of the Memex in
Support of Innovation. Technology and Culture, 61(2), 480–511.
https://doi.org/10.1353/tech.2020.0050
Lialina, O., & Espenschied, D. (2010). ONE TERABYTE OF KILOBYTE AGE [Web].
https://anthology.rhizome.org/one-terabyte-of-kilobyte-age
Lievrouw, L. A. (2014). Materiality and Media in Communication and Technology Studies: An
Unfinished Project. In K. A. Foot, P. J. Boczkowski, & T. Gillespie (Eds.), Media Technologies:
Essays on Communication, Materiality, and Society (pp. 21–51). MIT Press.
Liu, J., & Zhao, J. (2021). More than plain text: Censorship deletion in the Chinese social media.
Journal of the American Society for Information Science and Technology, 72(1), 18–31.
https://doi.org/10.1002/asi.24390
Lilley, A. (2006, March 20). Public is the new private for the MySpace generation. The Guardian.
https://www.theguardian.com/media/2006/mar/20/mondaymediasection.digitalmedia
Lipartito, K. (2003). Picturephone and the Information Age: The Social Meaning of Failure.
Technology and Culture, 44(1), 50–81. https://doi.org/10.1353/tech.2003.0033
List of defunct social networking services. (2022). In Wikipedia.
https://en.wikipedia.org/w/index.php?title=List_of_defunct_social_networking_services&old
id=1075222926

227
Lomas, N. (n.d.). MySpace Punishes Its Few Remaining Friends By Vanishing Their Blogs.
TechCrunch. Retrieved February 15, 2022, from
https://social.techcrunch.com/2013/06/12/bring-the-blogs-back/
Low, C. (2016, October 27). Twitter is killing Vine. Engadget. https://www.engadget.com/2016-10-
27-twitter-is-killing-vines-mobile-app.html
Lumen Learning. (n.d.). Technology and Innovation | Boundless Management. Retrieved April 24, 2022,
from https://courses.lumenlearning.com/boundless-management/chapter/technology-and-
innovation/
MacKenzie, D. (1996). Knowing Machines: Essays on Technological Change. MIT Press.
Martin, C. D. (1993). The Myth of the Awesome Thinking Machine. Communications of the ACM,
36(4), 120–133.
Marvin, C. (1988). When Old Technologies Were New: Thinking about Electric Communication in the Late
Nineteenth Century. Oxford University Press.
Marx, L. (2010). Technology: The Emergence of a Hazardous Concept. Technology and Culture, 51(3),
561–577.
Matthes, E. H. (2018). The Ethics of Cultural Heritage. In E. N. Zalta (Ed.), The Stanford Encyclopedia
of Philosophy (Fall 2018). Metaphysics Research Lab, Stanford University.
https://plato.stanford.edu/archives/fall2018/entries/ethics-cultural-heritage/
Mayer, V. (2008). Studying up and F**cking up: Ethnographic Interviewing in Production Studies.
Cinema Journal, 47(2), 141–148.
McBride, J., & Siripurapu, A. (2021). The State of U.S. Infrastructure. Council on Foreign Relations.
https://www.cfr.org/backgrounder/state-us-infrastructure

228
McElvaney, K. (2014, February 27). Agbogbloshie: The world’s largest e-waste dump – in pictures.
The Guardian. http://www.theguardian.com/environment/gallery/2014/feb/27/agbogbloshie-
worlds-largest-e-waste-dump-in-pictures
McLuhan, M. (1964). Understanding Media: The Extensions of Man. McGraw-Hill.
McRoberts, S., Ma, H., Hall, A., & Yarosh, S. (2017). Share First, Save Later: Performance of Self
through Snapchat Stories. Proceedings of the 2017 CHI Conference on Human Factors in Computing
Systems, 6902–6911. https://doi.org/10.1145/3025453.3025771
Medina, E. (2011). Cybernetic revolutionaries: Technology and politics in Allende’s Chile. MIT Press.
https://doi.org/10.7551/mitpress/8417.001.0001
Medina, E. (2018, October 14). Computers and the Struggle for Memory in Post-Pinochet Chile. SIGCIS:
Stored in Memory, St. Louis, Mo.
Meta. (2016, March 30). Engineering for nostalgia: Building a personalized “On This Day”
experience. Engineering at Meta. https://engineering.fb.com/2016/03/30/ml-
applications/engineering-for-nostalgia-building-a-personalized-on-this-day-experience/
Miller, C. C. (2011, June 28). Another Try by Google to Take On Facebook. The New York Times.
https://www.nytimes.com/2011/06/29/technology/29google.html
Milligan, I. (2015). Finding Community in the Ruins of GeoCities: Distantly Reading a Web Archive. 2.
Miltner, K. M., & Gerrard, Y. (2021). “Tom had us all doing front-end web development”: A
nostalgic (re)imagining of MySpace. Internet Histories, 0(0), 1–20.
https://doi.org/10.1080/24701475.2021.1985836
Mitchell, D. (2006, September 16). The Story Behind MySpace. The New York Times.
https://www.nytimes.com/2006/09/16/business/the-story-behind-myspace.html

229
Molina, J., & Guynn, B. (2016). Twitter slashes jobs, Vine as it seeks profits. USA TODAY.
https://www.usatoday.com/story/tech/news/2016/10/27/twitter-cuts-third-
quarter/92821316/
Montfort, N., & Bogost, I. (2009). Racing the Beam: The Atari Video Computer System. MIT Press.
Moore, M. H. (1995). Creating Public Value: Strategic Management in Government. Harvard University
Press.
Moreau, E. (2022, January 21). Is MySpace Dead or Does It Still Exist? https://www.lifewire.com/is-
myspace-dead-3486012
Morris, B. R. (2001, June 7). In Web Storage World, No Space at MySpace. The New York Times, 3.
Mosco, V. (2005). Digital Sublime: Myth, Power, and Cyberspace. The MIT Press.
Mudallal, Z. (2016, October 27). RIP vine: 10 vines you’ll keep on a loop forever: We did it for the
vine. The Washington Post.
Mulligan, B. (n.d.). MySpace Squandered the Only Thing It Had Left. TechCrunch. Retrieved February
15, 2022, from https://social.techcrunch.com/2013/02/02/myspace-squandered-the-only-
thing-it-had-left/
MySpace HQ. (n.d.). Foursquare. Retrieved March 7, 2022, from
https://foursquare.com/v/myspace-hq/4a65e0aff964a520adc71fe3
Nader, L. (1972). Up the anthropologist: Perspectives gained from studying up. In D. Hymes (Ed.),
Reinventing Anthropology (pp. 284–311). University of Michigan Press.
Natale, S. (2016). Unveiling the Biographies of Media: On the Role of Narratives, Anecdotes, and
Storytelling in the Construction of New Media’s Histories: Unveiling the Biographies of Media.
Communication Theory, 26(4), 431–449. https://doi.org/10.1111/comt.12099

230
Nieborg, D. B., & Helmond, A. (2019). The political economy of Facebook’s platformization in the
mobile ecosystem: Facebook Messenger as a platform instance. Media, Culture & Society, 41(2),
196–218. https://doi.org/10.1177/0163443718818384
Noble, S. U. (2018). Algorithms of Oppression. NYU Press.
Noble, S. U., & Tynes, B. (Eds.). (2016). The Intersectional Internet: Race, Sex, Class, and Culture Online.
Peter Lang.
Nora, P. (1989). Between Memory and History: Les Lieux de Memoire (M. Roudebush, Trans.).
Representations, 26(Memory and Counter-Memory), 7–24.
Nungesser, V.-S. (2009). Mediation, Remediation, and the Dynamics of Cultural Memory (A. Erll & A.
Rigney, Eds.; pp. 31–48). De Gruyter. https://doi.org/10.1515/9783110217384.1.31
Oakley, K., & O’Brien, D. (n.d.). Cultural Value and Inequality: A Critical Literature Review. 46.
O’Hara, K., Morris, R., Shadbolt, N., Hitch, G. J., Hall, W., & Beagrie, N. (2006). Memories for life:
A review of the science and technology. Journal of The Royal Society Interface, 3(8), 351–365.
https://doi.org/10.1098/rsif.2006.0125
O’Reilly, K. (2010). Combining sanitation and women’s participation in water supply: An example
from Rajasthan. Development in Practice, 20(1), 45–56.
https://doi.org/10.1080/09614520903436976
O’Reilly, K., & Dhanju, R. (2014). Public taps and private connections: The production of caste
distinction and common sense in a Rajasthan drinking water supply project. Transactions -
Institute of British Geographers (1965), 39(3), 373–386. https://doi.org/10.1111/tran.12039
O’Reilly, T. (2005, September 30). What Is Web 2.0. O’Reilly.
Orr, J. (1996). Talking about Machines: An Ethnography of a Modern Job. Cornell University Press.
Oxford English Dictionary. (n.d.-a). “salvage, n.”. (Oxford University Press). Oxford University Press
%C.

231
https://www.oed.com/view/Entry/170270?rskey=1zeV3c&result=1&isAdvanced=f
alse
Oxford English Dictionary. (n.d.-b). “value, n.”. (Oxford University Press). Oxford University Press
%C.
https://www.oed.com/view/Entry/221253?rskey=0Cu77M&result=1&isAdvanced
=false
Papacharissi, Z. (2015). We Have Always Been Social. Social Media + Society, 1(1),
2056305115581185. https://doi.org/10.1177/2056305115581185
Parsons, L. (2007). Thompson’s rubbish theory: Exploring the practices of value creation. ACR
European Advances.
Paßmann, J., & Schubert, C. (2021). Liking as taste making: Social media practices as generators of
aesthetic valuation and distinction. New Media & Society, 23(10), 2947–2963.
https://doi.org/10.1177/1461444820939458
Pastore, R. (2015, June 30). Mark Zuckerberg: The Future Of Facebook Is Sharing Thoughts. Fast
Company. https://www.fastcompany.com/3048091/mark-zuckerberg-the-future-of-facebook-
is-sharing-thoughts
Pearce, C. (2009). Communities of Play: Emergent Cultures in Multiplayer Games and Virtual Worlds. MIT
Press.
Perez, S., & Roof, K. (n.d.). Twitter is shutting down Vine. TechCrunch. Retrieved September 17,
2021, from https://social.techcrunch.com/2016/10/27/twitter-is-shutting-down-vine/
Peters, B. (2016). Digital. In B. Peters (Ed.), Digital Keywords (pp. 93–108). Princeton University
Press; JSTOR. https://doi.org/10.2307/j.ctvct0023.13
Piskorski, M. J., & Knoop, C.-I. (2006). Friendster (A). Harvard Business School.

232
Plantin, J.-C. (2018). Review Essay: How Platforms Shape Public Values and Public Discourse.
Media, Culture & Society, 41(2), 252–257.
Plantin, J.-C., Lagoze, C., Edwards, P. N., & Sandvig, C. (2018). Infrastructure studies meet platform
studies in the age of Google and Facebook. New Media & Society, 20(1), 293–310.
https://doi.org/10.1177/1461444816661553
Plantin, J.-C., & Punathambekar, A. (2019). Digital media infrastructures: Pipes, platforms, and
politics. Media, Culture & Society, 41(2), 163–174. https://doi.org/10.1177/0163443718818376
Pogue, D. (2011, October 19). Facebook Changes Inspire More Grumbling. The New York Times.
https://www.nytimes.com/2011/10/20/technology/personaltech/facebook-changes-inspire-
more-grumbling-state-of-the-art.html
Pollard, N. (2012). Alternative Histories in African-American Scrapbooks. In E. G. Garvey (Ed.),
Writing with Scissors: American Scrapbooks from the Civil War to the Harlem Renaissance (pp. 131–171).
Oxford University Press.
Porter, J. (2019, March 18). MySpace deleted 12 years’ worth of music in a botched server migration. The Verge.
https://www.theverge.com/2019/3/18/18271023/myspace-music-videos-deleted-2003-2015-
server-migration
President Biden’s Bipartisan Infrastructure Law. (n.d.). The White House. Retrieved March 23, 2022, from
https://www.whitehouse.gov/bipartisan-infrastructure-law/
Radford, G. P., Radford, M. L., & Lingel, J. (2012). Alternative libraries as discursive formations:
Reclaiming the voice of the deaccessioned book. Journal of Documentation, 68(2), 254–267.
https://doi.org/10.1108/00220411211209221
Raftery, B. (n.d.). RIP Vine, the Platform That Managed to Make 6 Seconds Feel Like Forever.
Wired. Retrieved December 7, 2021, from https://www.wired.com/2016/10/rip-vine/

233
Ramakrishnan, K., O’Reilly, K., & Budds, J. (2021). The temporal fragility of infrastructure:
Theorizing decay, maintenance, and repair. Environment and Planning. E, Nature and Space (Print),
4(3), 674–695. https://doi.org/10.1177/2514848620979712
Reads, 7 5k. (2006, September 12). MySpace: History of Spam & Deception Exposed. Search Engine
Journal. https://www.searchenginejournal.com/myspace-history-of-spam-deception-
exposed/3790/
Reier, S. (2000, June 10). Half a Century Later, Economist’s “Creative Destruction” Theory Is Apt
for the Internet Age: Schumpeter:The Prophet of Bust and Boom. The New York Times.
https://www.nytimes.com/2000/06/10/your-money/IHT-half-a-century-later-economists-
creative-destruction-theory-is.html
Reno, J., & Alexander, C. (2012). Economies of recycling: The global transformations of materials, values and
social relations. Zed Books.
Richardson, K., & Hessey, S. (2009). Archiving the self? Facebook as biography of social and
relational memory. Journal of Information, Communication and Ethics in Society, 7(1), 25–38.
https://doi.org/10.1108/14779960910938070
Roberts, S. T. (2019). Behind the Screen: Content Moderation in the Shadows of Social Media. Yale University
Press.
Roettgers, J. (2016, October 27). Twitter is shutting down vine. Variety.
http://variety.com/2016/digital/news/twitter-vine-shutting-down-1201902457/
Rogers, K. (2016, October 27). Vine Is Closing Down, and the Internet Can’t Stand It. The New York
Times. https://www.nytimes.com/2016/10/28/technology/vine-is-closing-down-and-the-
internet-cant-stand-it.html

234
Romano, A. (2016, October 28). You may not have understood Vine, but its demise is a huge
cultural loss. Vox. https://www.vox.com/2016/10/28/13439450/vine-shutdown-loss-to-
black-culture
Rosenberg, D. (2021). Data. In A. Blair, P. Duguid, A.-S. Goeing, & A. Grafton (Eds.), Information:
A Historical Companion (pp. 387–291). Princeton University Press.
Ross, R. (2003, September 8). Lifelog a useful, brutal reality checkLogged on for life; Technology
could soon record a person’s every moment in video, audio, and text What a memory-aid with
what consequences for humanity? By Rachel Ross: [ONT Edition]. Toronto Star. ProQuest
Central; ProQuest One Academic.
http://libproxy.usc.edu/login?url=https://www.proquest.com/newspapers/lifelog-useful-
brutal-reality-checklogged-on-life/docview/438629912/se-2?accountid=14749
Roy, S. (2011, June). Facebook Domination Vs MySpace Fall: Data Comparison. Online Marketing
Trends. http://www.onlinemarketing-trends.com/2011/06/facebook-domination-vs-myspace-
fall.html
Rozsa, M. (2016, October 27). Twitter announces it’s shutting down Vine, after announcing big
layoffs. Salon. https://www.salon.com/2016/10/27/twitter-announces-its-shutting-down-vine-
after-announcing-big-layoffs/
Sadowski, J. (2019). When data is capital: Datafication, accumulation, and extraction. Big Data
& Society, 6(1), 2053951718820549. https://doi.org/10.1177/2053951718820549
Safire, W. (1983, October 16). On Language; Dust Heaps of History. The New York Times.
https://www.nytimes.com/1983/10/16/magazine/on-language-dust-heaps-of-history.html
Safire, W. (2003, June 5). Opinion | Dear Darpa Diary. The New York Times.
https://www.nytimes.com/2003/06/05/opinion/dear-darpa-diary.html

235
Salimkhan, G., Manago, A. M., & Greenfield, P. M. (2010). The construction of the virtual self on
MySpace. Cyberpsychology: Journal of Psychosocial Research on Cyberspace, 4(1).
Sandvig, C. (2013). The Internet as Infrastructure. In W. H. Dutton (Ed.), The Oxford Handbook of
Internet Studies. Oxford University Press.
Schrage, M. (1995). At the Dawn of the Atomic Age, a Technocrat Ahead of His Time. The Los
Angeles Times.
Schumpeter, J. A. (1962). Capitalism, Socialism and Democracy. Harper Torchbooks.
Schwartz, J. M., & Cook, T. (2002). Archives, records, and power: The making of modern memory.
Archival Science, 2(1–2), 1–19.
Shea, M. (2019). Participatory Methods and Community-Engaged Practices for Collecting,
Presenting, and Representing Cultural Memory. In D. Drozdzewski & C. Birdsall (Eds.), Doing
Memory Research: New Methods and Approaches (pp. 169–189). Springer Singapore.
https://doi.org/10.1007/978-981-13-1411-7_9
Sheppard, S. N. (2017). I Love Cinema: Black Film and Speculative Practice in the Era of Online
Crowdfunding. Film Quarterly, 71(2), 25–31. https://doi.org/10.1525/fq.2017.71.2.25
Shilton, K. (2013). Values Levers: Building Ethics into Design. Science, Technology, & Human Values,
38(3), 374–397. https://doi.org/10.1177/0162243912436985
Shilton, K., & Greene, D. (2019). Linking Platforms, Practices, and Developer Ethics: Levers for
Privacy Discourse in Mobile Application Development. Journal of Business Ethics, 155(1), 131–
146. https://doi.org/10.1007/s10551-017-3504-8
Shilton, K., Koepfler, J. A., & Fleischmann, K. R. (2014). How to see values in social computing:
Methods for studying values dimensions. Proceedings of the 17th ACM Conference on Computer
Supported Cooperative Work & Social Computing, 426–435.
https://doi.org/10.1145/2531602.2531625

236
Shiozaki, R. (2021). Future uncertainties for preserving tweets: Peoples’ perceptions in Japan. Journal
of Librarianship and Information Science, 09610006211037392.
https://doi.org/10.1177/09610006211037392
Shontell, A. (2013, October 22). Vine Is The Fastest-Growing App Of 2013. Business Insider.
https://www.businessinsider.com/vine-is-the-fastest-growing-app-of-2013-2013-10
Sismondo, S. (1993). Some Social Constructions. Social Studies of Science, 23(3), 515–553.
https://doi.org/10.1177/0306312793023003004
Slade, G. (2007). Made to Break: Technology and Obsolescence in America. Harvard University Press.
Smit, R., Heinrich, A., & Broersma, M. (2017a). Witnessing in the new memory ecology: Memory
construction of the Syrian conflict on YouTube. New Media & Society, 19(2), 289–307.
https://doi.org/10.1177/1461444815604618
Smit, R., Heinrich, A., & Broersma, M. (2017b). Activating the past in the Ferguson protests:
Memory work, digital activism and the politics of platforms. New Media & Society, 20(9), 3119–
3139. https://doi.org/10.1177/1461444817741849
Smith, M. R. (1994). Technological Determinism in American Culture. In M. R. Smith & L. Marx
(Eds.), Does Technology Drive History?: The Dilemma of Technological Determinism (pp. 1–36). MIT
Press.
Smith, M. R., & Marx, L. (Eds.). (1994). Does Technology Drive History?: The Dilemma of Technological
Determinism. MIT Press.
Spurgeon, Christina. (2008). Advertising and new media. Routledge.
Staff, M. (2019, August 15). Here’s the Memo About the ‘Burden of Being Black at Google.’ Vice.
https://www.vice.com/en/article/zmjvy3/heres-the-memo-about-the-burden-of-being-black-
at-google

237
Star, S. L. (1999). The Ethnography of Infrastructure. American Behavioral Scientist, 43(3), 377–391.
https://doi.org/10.1177/00027649921955326
Star, S. L., & Bowker, G. C. (2010). How to Infrastructure. In L. A. Lievrouw & S. Livingstone
(Eds.), Handbook of New Media: Social Shaping and Social Consequences of ICTs (2nd ed., pp. 230–
245). SAGE.
Star, S. L., & Ruhleder, K. (1996). Steps Toward an Ecology of Infrastructure: Design and Access
for Large Information Spaces. Information Systems Research, 7(1), 111–134.
Stark, L., & Hoffmann, A. L. (2019). Data Is the New What? Popular Metaphors & Professional
Ethics in Emerging Data Culture. Journal of Cultural Analytics, 4(1), 11052.
https://doi.org/10.22148/16.036
Starosielski, N. (2015). The undersea network (First edition.). Duke University Press.
Steel, E. (2011, March 28). MySpace Decline Puts Off Advertisers. Wall Street Journal.
https://www.wsj.com/articles/SB10001424052748703576204576226620748953038
Steger, I. (n.d.). GeoCities Japan is finally shutting down. Quartz. Retrieved September 22, 2021, from
https://qz.com/1408120/yahoo-japan-is-shutting-down-its-website-hosting-service-geocities/
Sterne, J. (2014). Media Analysis Beyond Content. Journal of Visual Culture, 13(1), 100–103.
https://doi.org/10.1177/1470412913509457
Stobierski, T. (2020, November 12). What Are Network Effects? HBS Online Business Insights Blog.
https://online.hbs.edu/blog/post/what-are-network-effects
Stoler, A. L. (2002). Colonial Archives and the Arts of Governance: On the Content in the Form. In
C. Hamilton, V. Harris, J. Taylor, M. Pickover, G. Reid, & R. Saleh (Eds.), Refiguring the Archive
(pp. 83–102). Springer Netherlands. https://doi.org/10.1007/978-94-010-0570-8_7
Sturken, M. (1997). Tangled Memories: The Vietnam War, the AIDS Epidemic, and the Politics of
Remembering. University of California Press.

238
Sunset, v. (n.d.). In OED Online. Oxford University Press. Retrieved April 20, 2022, from
http://www.oed.com/view/Entry/55053068
Swartz, J. (2016, October 28). Vine won’t be last tech buy to fizzle. Reno Gazette, B4.
Tait, I. (2011, July 22). Android may give Google+ the edge in social media. Campaign.
Tandoc, E. C. (2014). Journalism is twerking? How web analytics is changing the process of
gatekeeping. New Media & Society, 16(4), 559–575. https://doi.org/10.1177/1461444814530541
The Internet’s first superservice is laid to rest. (2009, October 26). Mail Today. https://search-
proquest-
com.libproxy1.usc.edu/docview/238388102/AC6EA555693546EBPQ/71?accountid=14749
Thompson, Michael. (1979). Rubbish theory: The creation and destruction of value. Oxford University Press.
Tomer, A., George, C., Kane, J. W., & Bourne, A. (2021). America has an infrastructure bill. What
happens next? Brookings Institute. https://www.brookings.edu/blog/the-
avenue/2021/11/09/america-has-an-infrastructure-bill-what-happens-next/
Tossell, I. (2011, April 29). Notes on a social network dying young. The Globe and Mail.
https://www.theglobeandmail.com/technology/digital-culture/notes-on-a-social-network-
dying-young/article597732/
Trouillot, M.-R. (1995). Silencing the past. Beacon Press.
Tucker, C. (2019). Digital Data, Platforms and the Usual [Antitrust] Suspects: Network Effects,
Switching Costs, Essential Facility. Review of Industrial Organization, 54(4), 683–694.
https://doi.org/10.1007/s11151-019-09693-7
Tucker, I. (2019, April 27). What happens to our data when we die? Elaine Kasket on a digital
dilemma. The Observer.
Turner, Fred. (2006). From counterculture to cyberculture: Stewart Brand, the Whole Earth Network, and the rise
of digital utopianism. University of Chicago Press.

239
Turney, D. (2014, July 20). A digital archaeology dig. Sun Herald, 13.
Turns out, the Internet is not forever. (2019, December 22). Boston Globe, K2.
Twitter kills Vine but quarterly earnings look promising. (2016, October 28). Cyber Media / Athena
Information Solutions.
http://libproxy.usc.edu/login?url=https://www.proquest.com/magazines/twitterkills-vine-
quarterly-earnings-look/docview/1833112371/se-2?accountid=14749
Twitter To Discontinue Video-Sharing App Vine. (2016, October 28). NPR.
https://www.npr.org/2016/10/28/499710864/twitter-to-discontinue-video-sharing-app-vine
Ulunma. (2020, March 20). Before Facebook there was… Friendster? Yes, that’s right! HBS Digital
Innovation and Transformation. https://digital.hbs.edu/platform-digit/submission/before-
facebook-there-was-friendster-yes-thats-right/
Van Dijck, J. (2007). Mediated Memories in the Digital Age. Stanford University Press.
van Dijck, J. (2009). Users like you? Theorizing agency in user-generated content. Media, Culture &
Society, 31(1), 41–58. https://doi.org/10.1177/0163443708098245
Van Dijck, J., Poell, T., & de Waal, M. (2018). The Platform Society: Public Values in a Connective World.
Oxford University Press.
van Duyn, A., & Waters, R. (2006, August 7). Google in $900m ad deal with MySpace. Financial
Times. https://www.ft.com/content/17e8e67e-2660-11db-afa1-0000779e2340
Van House, N., & Churchill, E. F. (2008). Technologies of memory: Key issues and critical
perspectives. Memory Studies, 1(3), 295–310. https://doi.org/10.1177/1750698008093795
Video: The Fall of Friendster. (2006, October 14). New York Times.
https://www.nytimes.com/video/business/1194817120397/the-fall-of-friendster.html
Villa-Nicholas, M. (2019). Latinx Digital Memory: Identity Making in Real Time. Social Media +
Society, 5(4), 2056305119862643. https://doi.org/10.1177/2056305119862643

240
Vine FAQs. (n.d.). Retrieved April 9, 2022, from https://help.twitter.com/en/using-twitter/vine-
faqs
Vinsel, L. (2014a, September 22). How to Give Up the I-Word, Pt. 1. Culture Digitally.
http://culturedigitally.org/2014/09/how-to-give-up-the-i-word-pt-1/
Vinsel, L. (2014b, September 23). How to Give Up the I-Word, Pt. 2. Culture Digitally.
http://culturedigitally.org/2014/09/how-to-give-up-the-i-word-pt-2/
Vinsel, L., & Russell, A. L. (2020). The innovation delusion (First edition.). Currency.
Wardrip-Fruin, N., & Montfort, N. (2003). The New Media Reader (Vol. 1). MIT press.
West, K. (2016). Famous: Vol. The Life of Pablo. GOOD Music/Def Jam Recordings.
West, S. M. (2019). Data Capitalism: Redefining the Logics of Surveillance and Privacy. Business &
Society, 58(1), 20–41. https://doi.org/10.1177/0007650317718185
Where Is All My Old Stuff? (n.d.). Help Center. Retrieved April 4, 2022, from
https://help.myspace.com/hc/en-us/articles/202233310-Where-Is-All-My-Old-Stuff-
Why did DARPA cancel LifeLog on the same day Facebook was launched? (n.d.). Quora. Retrieved January
18, 2022, from https://www.quora.com/Why-did-DARPA-cancel-LifeLog-on-the-same-day-
Facebook-was-launched
Williams, R. (1976). Keywords: A Vocabulary of Culture and Society. Croom Helm.
Wilson, K. (2011, July 13). Kevin Wilson column. McClatchy - Tribune Business News.
Winner, L. (Ed.). (1980). Do Artifacts Have Politics? Daedelus, 109(1).
https://doi.org/10.4324/9781315259697
Wortham, J. (2011a, March 8). Memolane Transforms Social Data Into Visual Timelines. The New
York Times. https://bits.blogs.nytimes.com/2011/03/08/memolane-transforms-social-data-
into-visual-timelines/?searchResultPosition=44

241
Wortham, J. (2011b, April 26). Friendster to Erase Early Posts and Old Photos. New York Times.
https://search-proquest-
com.libproxy1.usc.edu/docview/2217182171/1AB52A96CE4A4A30PQ/4?accountid=14749
Wunker, S. (n.d.). Four Morals From MySpace’s Fall. Forbes. Retrieved April 4, 2022, from
https://www.forbes.com/sites/stephenwunker/2011/07/25/4-morals-from-myspaces-fall/
Wylie, A. (2005). The Promise and Perils of an Ethic of Stewardship. In L. Meskell & P. Pels (Eds.),
Embedding Ethics (1st ed., pp. 47–68). Bloomsbury Academic; Bloomsbury Collections.
http://www.bloomsburycollections.com/book///
Yalcinkaya. (2022, February 28). Why are we all so obsessed with early web nostalgia? Dazed.
https://www.dazeddigital.com/science-tech/article/55404/1/why-are-we-all-so-obsessed-
with-web-1-nostalgia
Yates, F. A. (2013). Art Of Memory. Taylor and Francis. https://doi.org/10.4324/9781315010960
Yeo, R. (2021). Notebooks. In A. Blair, P. Duguid, A.-S. Goeing, & A. Grafton (Eds.), Information: A
Historical Companion (pp. 636–640). Princeton University Press.
Zachary, G. Pascal. (1997). Endless frontier: Vannevar Bush, engineer of the American Century. Free Press.
Zelizer, B. (2002). Finding aids to the past: Bearing personal witness to traumatic public events.
Media, Culture & Society, 18.
Zimmer, M. (2008). Preface: Critical Perspectives on Web 2.0. First Monday, 13(3).
https://firstmonday.org/ojs/index.php/fm/article/download/2137/1943
Zittrain, J. L. (2014). Reflections on Internet Culture. Journal of Visual Culture, 13(3), 388–394.
https://doi.org/10.1177/1470412914544540

242
Appendix A: Methods
This dissertation used qualitative social scientific and interpretive approaches to investigate
the process of sunsetting. I describe these methods in detail here, first chronicling the rationale for
choosing the platforms that were analyzed in this project, then discussing methods of data gathering
and analysis, including both the document analysis that informed Chapters 1 and 5 and the
interview-based analysis that informed Chapters 2, 3 and 4. I conclude by reflecting on how the
COVID-19 pandemic, which struck just as I was beginning empirical research, shaped this project.
Choice of Platforms
In the Introduction, I noted that platform sunsetting was not a rare occurrence. If this is the
case, how does one choose among the multitude of social media platforms that have sunset? This
was one of the earliest methodological problems I faced. Before choosing which platforms to focus
on, however, I needed to also decide if I was indeed focusing on platforms instead of an individual
platform. While dedicating all research efforts toward one platform would have allowed for a
potentially more comprehensive understanding of one sunsetting process, it could risk the ability to
draw conclusions about the process of platform sunsetting more generally. As such, I decided to
analyze multiple platforms.
From there, I focused on platforms based in the United States. This was for two reasons:
one, there was the practical consideration of encountering few linguistic barriers when attempting to
interview employees whose workplaces were primarily English speaking. The second is that US-
based platforms are often the sites that are dominant networks in other countries as well, as can be
seen in Friendster’s case. Moreover, US-based, and especially Silicon Valley-based, techno-cultures
tend to, in the words of Fred Turner (2016, p. 257), “export their modes of thinking and working far
beyond the confines of Silicon Valley,” a sort of travelling hegemony that makes these spaces
particularly powerful ones to look at.

243
I decided I would begin with ‘culturally significant’ social media platforms in this social
context, knowing that factors like access and representativeness may cause me to look at platforms
that were smaller, ones that were less likely to be household names. Culturally significant, in this
case, was a qualitative descriptor that was arrived at through metrics like number of users, press
attention, formerly large economic valuations, and formerly high-ranking on charts monitoring
website traffic, like Alexa or ComScore. I chose four platforms to start with: GeoCities, Friendster,
MySpace, and Vine, representing slightly different era and foci, but all considered social media
platforms under the definitions I outlined in the Introduction.
While I was able to recruit participants successfully from these platforms (see Participant
Identification and Outreach section below), I seemed to rarely be able to access anyone who had
been at a platform at the very end, especially technical employees who could describe the process of
removing user data. Indeed, as I found out through these interviews, this was sheer probability:
when a platform shuts down, there are typically so few people left at the company that the chances
of locating employees who were there and then having them agree to be interviewed were quite slim.
Because of this, as I continued interviews with employees from large platforms, I also decided to
start looking at smaller platforms, reasoning that I could likely recruit former founders and CEOs
who likely had closer involvement with closure and could speak to the platform’s sunsetting in
detail. A Wikipedia page chronicling defunct platforms (‘List of defunct,’ 2022), as well as the
Archive Team’s listing of closed platforms (‘Deathwatch,’ 2022), were both useful in finding
additional platforms to look at. While interviews with leaders from these smaller platform companies
did not end up being the main analytic focus of the work shown here, they did provide valuable
background for understanding processes of shutdown in greater detail.

244
Document Analysis: Chapters 1 and 5
Both chapters 1 and 5 relied on document analysis, or the “systematic procedure for
reviewing or evaluating documents” which consists of “finding, selecting, appraising (making sense
of), and synthesizing data contained in documents” (Bowen, 2009, p. 27). Chapter 1 took an
interpretive approach to document analysis to trace the construction of digital memory
infrastructure. It combined the analysis of primary sources like grey material, presentation videos,
and press pieces, with the analysis of secondary sources like scholarly literature on memory and
media.
Chapter 5 approached document analysis using grounded theory to develop the concept of
infrastructural expectations. First, I found and selected documents for analysis. Using the database
ProQuest, I gathered English-language articles that included terms related to sunsetting, like
“sunset,” “closure,” “defunct,” “delete,” in combination with the platform of interest, whether
GeoCities, Friendster, MySpace, or Vine. I focused this search on the dates of sunsetting events as
outlined in Chapter 5. While the bulk of my collection came from searching near these dates, I also
removed temporal filters on ProQuest to make sure that articles on the subject were captured even
if published after the sunsetting event. These searches were then conducted in Google News for
additional coverage. In total, I analyzed 252 articles, the breakdown of which are in Chapter 5.
To make sense of these media items, I used qualitative data analysis software NVivo to
perform document analysis using a grounded theory approach. As Strauss and Corbin (1990, p. 23)
write, “grounded theory is one that is inductively derived from the study of the phenomenon it
represents. That is, it is discovered, developed, and provisionally verified through systematic data
collection and analysis of data pertaining to that phenomenon.” Because I had little concerted
awareness of major themes in press covering sunsetting, nor was there literature on this
phenomenon, grounded theory was useful in generating novel themes from the gathered data. Using

245
the “constant comparative method” associated with grounded theory (Glaser and Strauss, 1967) I
checked and compared data, clustered subthemes emerging from this data into broader categories,
and compared these categories across all press items, later synthesizing this data into the analysis that
makes up Chapter 5.
Interviews and Thematic Analysis: Chapters 2, 3, and 4
Chapters 2, 3, and 4, drew on analyses of 56 interviews I conducted between March of 2020
and January of 2021. Fifty-three of these interviews were with former employees of platforms. The
breakdown of platforms by number of participants can be seen in the Introduction, Table 1. Three
interviews were with people who held positions related to sunsetting and data loss, including two
employees of archival organizations and one Google employee who had built a tool for exporting
user-generated content. A breakdown of these interviews, including name (typically as a
pseudonym), the organization they were affiliated with that was of interest to this project, their
occupation, the mode of interview, and whether excerpts from their interview were included in this
writing, can be seen in Table 3. IRB approval for these interviews was received from the University
of Southern California’s Office for the Protection of Research Subjects on January 10, 2020.
Participant Identification and Outreach
Having chosen the main platforms of interest for this project, as outlined in the
Introduction, I identified potential participants through business social network LinkedIn. Using a
paid subscription, I was able to search for participants who had previously worked at a given
platform. While I often focused my outreach on those who worked there actively during a
platform’s dismantling process–for instance, working at the platform during the time of data
deletion–I also cast a wider temporal net. Because the dismantling process extends beyond the
moment of data loss, and sunsetting is shaped by the values, expectations, and decisions made prior
to and following data deletion, I also engaged in outreach to employees who worked at the

246
platforms of interest prior to the sunsetting era. This recruitment process would often shift as I
began interviews for a particular platform, based on filling in information for eras or positions that
were not yet well represented in the collected data.
On LinkedIn, I messaged individuals that I had aggregated on individual platform lists with a
request for an interview. The outreach message included information on who I was, “a PhD
candidate in communication at the University of Southern California;” how I found them; a brief
description of the dissertation project, which I typically phrased as the “entire lifecycle of social
media platforms;” the types of topics we would cover, including what they did while working there,
how the platform changed, what people envisioned for the platform over time; and how the
interview would be conducted. For those who had publicly available contact information, I would
send an email instead of a message on LinkedIn, as the subscription model limited the number of
messages that could be sent per month. I recruited until I reached saturation, a point I discuss in
more detail later in the Data Saturation & Triangulation section. I also had limited success snowball
sampling, with two direct introductions made by a participant and five coming through names that a
participant gave me to reach out to. These introductions were more likely to happen for platforms
that existed more recently (e.g. Vine, MySpace); employees who worked on platforms longer ago
(e.g. Friendster, GeoCities) often felt they no longer had strong connections to their previous
coworkers.
I estimate I heard back from between 10 and 15% of the people I contacted. Fittingly, given
this project’s subject, I would have been able to calculate this figure more precisely had I retained
access to the messages I sent. However, LinkedIn removes access to all previous messages sent
while using their premium service upon termination of the contract, a feature that was not readily
apparent to this customer.

247
Participant Demographics and Positionality
While I did not ask my participants for demographic information, there are general
demographic themes that I can speak to. My reading of this demographic information inevitably
relies on assumptions about gender, race, and age, and is meant to give a general picture to the
reader, with the acknowledgement that my assumptions may very well be incorrect. With that in
mind, most of my participants presented as men (apprx. 46 of 56 participants; 82%), with the large
majority also presenting as white (apprx. 40 of 56 participants; 71%). Participants of color presented
as Black or as having ethnic backgrounds in East and Southeast Asia. All but two participants
resided in the United States. They appeared to range in age from about 30 to their mid-50s. These
approximations are generally in line with the demographics of the U.S. high-tech industry as it stands
today, with demographics in this project skewing slightly whiter and more male (see Harrison, 2019).
Age also reflected industry demographics, with a relatively consistent range that put a participant’s
age while working at a platform typically in their mid-20s to early 30s.
My own subject-position–white, female, middle class, on the younger side of middle-age,
American and highly educated–undoubtedly affected this project. While the US tech industry,
especially at the executive level and in technical roles, is dominated by men (Harrison, 2019), I did
not necessarily feel my status as a woman positioned me as an outsider, especially given the context
of a one-on-one interview. That is, any outsider status may have been more readily felt in an large
industry setting where my standing as a graduate student social scientist was not known (e.g.
Dunbar-Hester, 2020 p. 185). Indeed, my status as a (female, white) PhD student afforded me the
chance to be a curious outside observer, a role that seemed legible to my participants, who seemed
prepared for me to ask follow-up questions or to ask them to explain unfamiliar procedures in detail.
However, while studying for a PhD at a well-known university affords some amount of
social status, I was generally “studying up,” meaning that I was looking upward on a chain of social

248
hierarchy (Nader, 1972), with participants affiliated with elite institutions and possessing high socio-
cultural (and socio-economic) status. There are unique challenges whether studying ‘up,’ ‘down, or
‘sideways;’ typically the most well-known challenge in studying up is the challenge in access. I was
expecting these challenges, and the 10 to 15% rate of response alludes to them. For particularly
high-status participants, access was often restricted because they set forth a limited period of time in
which our interview could occur (i.e. 30 minutes instead of 1 hour). I also noticed that these higher-
status participants would often steer the conversation more strongly, leaving less room for me to ask
my prepared questions. In these interviews, while I often tried to redirect, I was sometimes ‘along
for the ride’ based on what the interviewee wanted to talk about.
Of course, there were forms of privilege and discrimination that were less readily visible to
me, and I’m grateful for colleagues for engaging me on these points. A major one is access to
participants. In media scholar Vicki Mayer’s (2008, p. 143) essay reflecting on shifts in cinema
studies from audience studies to production studies, she notes how the social status of the
researcher shapes access to elites working at the production-level of these systems:
Entry and access to different production worlds seems to depend very much on who we are,
the social worlds we inhabit, and the positions in the workplace hierarchy we have in
academia. Certain scholars have marveled at the ease of access to producers, while others
find themselves unable to enter this world.
While situated within cinema studies, the observation can be mapped to the production-level
relations I was examining on social media platforms. As I engaged in recruitment, I was surprised at
the number of people who were willing to talk to me–and it is worth being critical about that relative
success. Colleagues have reminded me that women of color recruiting with the same methods that I
used would very likely be ‘read’ by potential interviewees differently, particularly as investigating the
tech industry with an assumed critical stance. In contrast, the social categorization of whiteness as

249
unmarked and normative, not to mention associations of white femininity as innocent and morally
righteous (Dyer, 2008), may have presented me as a ‘non-threatening’ interviewer, especially to the
tech industry’s major demographic–white men–shaping who I was able to interview and how those
interviews proceeded.
Finally, my status as an American, who chose to interview employees from platforms almost
exclusively based in the United States, inevitably gives this project a limited perspective. This is not
only the limited perspective of the US tech industry, but also the limited perspective on memory
practices, as I emphasize (especially in Chapter 1) North American and European mnemonic
traditions that employ documentation and mediation in culturally specific ways.
Questions and Interview Procedure
I developed questions that attempted to elicit descriptions of work at a platform, what that
work looked like over time, and how one proceeded specifically through the process of sunsetting.
To elicit answers related to the broad concepts of value and values that I was interested in, I chose
open-ended questions where these matters would be addressed implicitly. That is, I did not develop
questions like: ‘What were your personal values as you shut the platform down?’ which would likely
prompt respondents to answer in ways that were pro-social if not accurate, for instance, ‘I valued
honesty and goodwill.’ Instead, I asked questions like: ‘Can you walk me through how you made that
decision?,’ or ‘If you could have designed the sunset how you wanted, what would that have looked
like?’ Questions also addressed the individual history of each platform.
Interviews were semi-structured, meaning that I had a series of questions I prepared, but
during conversation I asked follow-ups, generated new questions, and took new directions based on
the information the interviewee was providing. I began interviews by introducing myself, the project,
asking for permission to record audio of the interview, confirming their choice for use of a real
name, pseudonym, or complete anonymity in the written version of this project (see ‘Anonymity and

250
Other Ethical Considerations’ section), and if they had any questions for me as we began. (For the
limited number of interviews who requested no audio recording, I instead took notes with the
interviewee’s permission.) The interview questions covered a participant’s role at the company, their
perception of user communities, the arc of the platform while they worked there, their perception of
the platform’s legacy, and their involvement with the platform’s closure, if applicable. A generic set
of questions is included in Appendix B. Interviews typically lasted between 45 minutes and an hour.
The shortest interview was 30 minutes, and the longest two hours.
Analysis
Interviews that I audio-recorded were transcribed using a third-party service; some
interviews were transcribed using a natural-language processing algorithm that transcribed
automatically, others were done by professional transcribers. I edited interview transcripts in either
case. Transcripts were imported into qualitative data analysis software NVivo, which I used for
qualitative thematic analysis.
Qualitative thematic analysis refers to the flexible method for “identifying, analysing and
reporting patterns (themes) within data,” typically interpreting aspects of the broader research topic
(Braun & Clarke, 2006, p. 79). Thematic analysis’ open-ended flexibility prompts analysts to make
decisions informed by a general theoretical framework; this project moves forward with a
‘contextualist’ approach, retaining a focus on the material limits of reality while interpreting how
individuals make meaning of their experiences within broader social contexts (ibid). Because this
project at once told the story of platforms over time (a realist approach) while analyzing the ways
that meaning, value, and values were articulated within these stories (constructivist), the hybrid form
of contextualism was deemed most appropriate.
Moreover, qualitative thematic analysis relies on either or both deductive and inductive
coding schemas, the former referring to an approach where themes are informed primarily by related

251
literatures and the latter referring to a grounded approach where themes emerge primarily from the
data. Acknowledging that I was working with literatures that had consistent themes but that I was
exploring a topic that had little academic scholarship dedicated to it, I chose to use a hybrid
deductive-inductive coding schema. Interviews were first analyzed using codes developed from
literature on digital memory, platforms, and infrastructural breakdown (e.g. technical obsolescence),
and then analyzed with another round of open coding to capture additional concepts (e.g.
responsibility for UGC). Interviews were coded iteratively following these two rounds, moving back
and forth between interviews from different platforms.
Data Saturation & Triangulation
There were several practical constraints that helped determine the January 2021 endpoint for
recruiting and interviewing participants, including limited time and dwindling access to additional
research funding. The most important factor, however, was in reaching a point of data saturation, a
moment when the themes that were emerging in my interviews were largely redundant, and new
themes were limited. These are both signals for bringing data collection to an end.
At the same time, I encountered a grey area in determining when data saturation was
reached. Because this project wove together individual histories of different platforms from diverse
eras with themes that arose across platforms, the point of saturation was less clear-cut than in other
projects with a more consistent participant population. For instance, while I may have heard a
significant amount about ‘infrastructural cost versus revenue’ considerations from employees at
GeoCities, Friendster and MySpace–making this theme highly redundant across platforms–had I
reached saturation if I had not yet encountered it multiple times in my interviews with Vine
employees? Would I be able to compare platforms if I had not reached saturation with this theme
for all platforms? (This is just an example: ‘infrastructural cost versus revenue’ was a major
consideration for Vine as well.)

252
I navigated this grey area through a couple tactics. For one, I added or modified questions to
make sure that interviewees had the chance to address these themes if they were present. At the
same time, I paid attention to ‘silences’ in themes across these platforms. For instance, if
‘infrastructural cost versus revenue’ was brought up by interviewees numerous times at other
platforms but it did not surface in several interviews with Vine employees, these lacunae would
signal that perhaps Vine did not experience comparable economic distress, or that it had put in place
technical infrastructure in an era when it was cheaper and more efficient; these are both theories that
would then be investigated through iterative questions.
While data saturation addresses empirical comprehensiveness, data triangulation addresses
empirical accuracy. A challenge in doing recent historical research through interviews is that data
gathering relies on interviewees’ ability to recall events that, in this project, happened between 25
and five years ago. To ensure accuracy as I synthesized a narrative from these stories, I triangulated
individuals’ statements with press accounts and the accounts of other participants. If I could not
adequately triangulate data, I instead noted the contested nature of a claim within the text. At the
same time, in line with the contextualist approach to thematic analysis discussed in the previous
section as well as memory studies’ movement from a focus on mnemonic accuracy to processes of
remembering and forgetting (Shea, 2018), I paid attention not just to the precision of the stories but
how employees recounted these stories, for instance, which values they emphasized or what issues
were especially salient to them.
Moreover, focusing on the construction of memory through media as a subject while relying
on the spoken memories of former employees (which sometimes involved the elicitation of media)
produced generative overlaps. For instance, this dissertation observes that communities and
individuals use media from the past to structure and strengthen identity in the present; so too did
the stories participants recounted relate to identity construction in the present (Drozdzewski &

253
Birdsall, 2018). An article drawing on interviews for this project with Friendster employees, for
instance, explores how in interviews participants would recount what they believed caused the
‘death’ of that platform, a reflective diagnosis that attested to their industry knowhow (Corry, 2021).
Anonymity and Other Ethical Considerations
When I set out to interview employees at platform companies, I imagined they would be
hesitant to talk with me. Employees from extant social media platforms, after all, are infamously
cagey about their work (e.g. Roberts, 2019); anonymous memos from former employees are often
the most unvarnished glance the public gets into their major tech companies inner workings (e.g.
Motherboard Staff, 2019). Often, these employees have signed non-disclosure agreements (NDAs)
restricting what they can say about their companies, even after employment ended. Moreover, they
may have been trepidatious that a researcher would cast a critical eye on their employment: public
critiques of the socio-political roles of large technology companies and social media platforms
appeared often over the course of 2020 and 2021, whether the Netflix documentary The Social
Dilemma or the anti-trust hearings where members of the US House of Representatives grilled the
CEOs of large tech companies.
This is all to say that those employees who agreed to be interviewed I expected would want
either pseudo-anonymity or complete anonymity, and I began our interviews by talking about the
level of anonymization they would like in any public version of this work, from articles to
presentations to the dissertation itself. A surprising number of interviewees said they were
comfortable using their real names: 23 of the 56 people I interviewed, or 41%. Granted, a significant
number of these were either relatively public figures (for instance, the head of archiving at the
Internet Archive), or were founders of a platform company; moreover, pseudonymity would not do
much if the real name of their platform was also being used in conjunction with their title.

254
Still, several platform employees also were comfortable with the use of their real name.
However, while writing this up, I chose to give every person I interviewed–except for public figures,
company founders, and company CEOs who agreed to use of a real name–a pseudonym. Ultimately,
I believed that using the real names of some employees and pseudonyms for others could threaten
the desired privacy of those participants who had requested pseudonyms. Sometimes employees
worked on small teams, and naming a few of them with real names and a few with pseudonyms
could potentially allow a reader to deduce who the pseudonymized employees were. For those who
requested a pseudonym, I also abstracted their position at the company in writing and in Table 3; for
instance, a ‘senior database specialist’ may become a ‘database administrator,’ a role occupied by
multiple people at a given company. I chose pseudonyms that reflected attributes of employees’ own
names without referencing those names; for instance, someone with a men’s name popular in the US
in the mid-1980s–for instance, Scott–would be renamed with a men’s name that was also popular
during this era–for instance, Bryan. As I discussed in Chapter 3, some employees emphasized their
desire for pseudonymity at instances in our conversation, either for reasons of personal privacy or
because they were afraid that some aspects of what they were saying could risk their future
employment. In these cases, I chose to not quote what I deemed to be more sensitive or identifiable
comments made by these employees.
One potentially risky hurdle was the NDA: a few employees mused–in all cases, somewhat
facetiously–about whether they had signed an NDA during their time working for the platform. At
these points, I reiterated that they did not need to share any information with me that they did not
wish to divulge, and that they could be on background if they did want to share information without
it being associated with a name or pseudonym. While NDAs are binding legal contracts that
technically can be enforced even after the dissolution of a company, the risks of this happening are
minimal as these companies either don’t exist at all or have almost no resources dedicated to them.

255
Nonetheless, I attempted to minimize these risks as much as possible through a combination of
pseudonymization and employing information on background.
Effect of COVID-19 Crisis on Methods
I had planned the first interview for this project to take place at a café near an interviewee’s
home on the west side of Los Angeles on March 24, 2020. We had confirmed the date, time and
location after my initial outreach on March 3
rd
. On March 12, the World Health Organization
declared COVID-19 a pandemic. On March 19, the governor of California had put in place a “stay-
at-home” order. USC had moved classes online, including the one I was teaching, and any research-
related travel was suspended. In the three weeks between my initial outreach and the date of our
planned interview, much had changed about the world and too about how I planned to gather data
for this project. The interview went forward on March 24, but like most ‘social’ activities, it would
be conducted remotely, in this case, over Zoom. Nearly all my interviews would be, except for those
I conducted over phone or via email.
While the primary empirical portion of this project was always envisioned as interview-
based, prior to the onset of COVID-19 I had planned to do the bulk of these interviews in person,
travelling to the main cities where my interviewees were located, including Los Angeles, San
Francisco, and New York City. I believed that in-person interviews in locations that are tech centers
like these had distinct advantages, allowing for a degree of participant observation that could
supplement the interviews and add color to the story of sunsetting. While I still believe that
participant observation would have added to this project, I am also convinced that I was able to
interview a greater number of people because I conducted the interviews virtually. For one, because
I did not travel, I was able to spend less time transcribing interviews as I had more funds available to
have them transcribed by a third-party. This freed up the time and energy to conduct more
interviews. Moreover, many of my participants likely had greater flexibility to fit an interview into

256
their day. To my knowledge, all my interviewees held occupations that were easily transferable to
working in a remote, online environment. Most joined the interview from what appeared to be their
homes and were likely afforded somewhat more flexibility to move between a work meeting and an
interview than they would have had in an in-person environment. The greater number of interviews
I was able to conduct holds its own advantages, including more perspectives and a better ability to
triangulate between different people’s stories.
I also wondered how the crisis had affected the demographics of my participant group, as
well as how the crisis had shaped the interviews themselves. As noted before, the vast majority of
my interviewees presented as men. This is undoubtedly caused by the significant male-skew in the
tech industries, but it also may have been influenced by the disproportionate caretaking burden that
women were shown to have taken on during COVID-19, including caring for children who
otherwise would have been at a daycare or school. It would make sense if those who had taken on
more caretaking duties would be less able to accommodate an interview.
Moreover, the stresses of the pandemic may have affected these interviews in ways that are
impossible to know, as people encountered grief, illness, uncertainty, and turmoil. For instance, I
recall interviewing someone who now worked in an executive role at a major technology company.
He was animated in our interview but was also constantly interrupted by Slack messages and emails.
As he was, he alluded to a crisis of sorts going on at his company. By happenstance, the next day I
read an article which noted that his company had just laid off a significant number of their
employees in reaction to the COVID-19 crisis, a task this interviewee would have been playing a
major role in the day of our conversation. This is only the most visible example–and one that did
not directly threaten the livelihood of this particular employee, just the livelihood of his own
employees–but it nonetheless signals the shifting ground my participants had to walk on.

257
In sum, COVID-19 significantly affected how this project was carried out, in ways both
clear–for instance, switching to interviews over Zoom instead of in-person interviews–and in ways
that are harder to define.
Table 3. Interviewees
Name Workplace Role at workplace Format Quoted?
1. Wallace GeoCities Community manager Zoom *
2. Patrick GeoCities Business administrator Zoom *
3. Maya GeoCities Volunteer community moderator Zoom
4. Zach GeoCities Business strategy director Zoom
5. Jordan GeoCities Legal Zoom
6. Andy GeoCities Developer Zoom
7. Scott Friendster CEO Zoom *
8. Richard Friendster CEO Zoom *
9. Sivu Friendster CTO Zoom *
10. Peter Friendster Technical architect Zoom *
11. Tom Friendster Developer Zoom *
12. Steve Friendster Int’l. development manager Phone *
13. Fidelia Friendster Product manager Zoom *
14. Lee Friendster Technical architect Phone *
15. Kelly Friendster Product manager Phone *
16. Jerome Friendster Developer Zoom *
17. Billy Friendster Database architect Zoom *
18. Liam Friendster Operations & IT Zoom
19. Riley Friendster Developer Zoom
20. Brendan MySpace Project manager Phone *

258
21. Leah MySpace Music manager Zoom *
22. Anonymous MySpace Anonymous Email
23. Gavin MySpace Product manager Zoom *
24. Troy MySpace Database administrator Zoom *
25. Lydia MySpace Content producer Zoom *
26. Michael MySpace Content producer Phone *
27. Anonymous MySpace Anonymous Zoom *
28. Albert MySpace Developer Zoom *
29. Doug MySpace Developer Zoom *
30. David MySpace Creative content director Zoom *
31. Krish MySpace Developer Phone *
32. Dev MySpace Technology finance Phone *
33. Louis MySpace Database administrator Zoom *
34. Julian MySpace Writer Zoom *
35. Aaron MySpace eUniverse board member Zoom *
36. Anonymous MySpace Anonymous Phone *
37. Nick MySpace Business development Zoom
38. Oliver MySpace Developer Zoom
39. Anonymous MySpace Anonymous Zoom *
40. Drew Vine Developer Zoom *
41. Corey Vine Creator manager Zoom
42. Allison Vine Content editor Zoom *
43. Caroline Vine Marketing Zoom *
44. Austin Vine Music editor Zoom *
45. Anonymous Vine Anonymous Email

259
46. Paul Vine Designer Zoom *
47. Becky Vine Creator representative Zoom *
48. Gibby Makeout Club Founder Zoom *
49. Brian Google Takeout Creator Zoom
50. Jane Bolt Founder Zoom
51. Dan Bolt Founder Zoom
52. Michael Couple Founder Zoom *
53. Josh 43Things Founder Zoom
54. Dale Friends Reunited Product manager Zoom
55. Jefferson Internet Archive Director of archiving Zoom
56. Jason Archive Team Founder Phone

260
Appendix B: Sample Interview Questions

Basic information

• Why did you choose to work there? How was the job pitched to you?
• What was your perception coming in of X’s place in the cultural zeitgeist?
• Had you been a user of X?
• What did the [person’s role at company] job entail?
• What was the arc of the company during your time there?

Team and office

• How was the workplace structured at that time?
• Who did you work with?
• Where was the office?
• What was the general atmosphere of the X office at that time?

General and changes to site

• During your time that you worked there, where did people imagine that X was going?
• What did people imagine 1 year, 5 years, 10 years down the line?
• How was that planned for in your role?
• Where did you think it was going?
• Did that change over time at all?
• Were there any major changes to the site while you worked there?
• How did users take those changes?
• How did X as a company think about its users at that time?
• (How) did concerns about monetization/making money factor into your work?
• What were some of the big dilemmas you faced?

Sunsetting

• In the intervening years, did you follow what was going on with X at all?
• Was there ever any discussion that the platform would ever shut down?
• Can you walk me through the X sunsetting process from when the thought first crossed
your mind that X would have to shut down?
• How does someone in your role encounter the process of winding a platform down?
• What steps are involved with shutting down a platform?
• What were challenges you encountered as you shut down X?
• Do these things usually go as planned?
• Is it a slow process or a quick process?
• Was it emotional to shut down the platform?
• How did you think about users at this time?

261
• Is there a time when you thought X could have been declared a “dead” platform?
• What’s your opinion on how X shut down? Is this a good sunset, a bad sunset?
• What would an ideal sunset look like to you? If you shut off X in the right way, what would
that have looked like for you?

After sunsetting
• What do you think about X’s overall legacy?
• Anything I should have asked about that I didn’t?

Abstract (if available)

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Cryptographic imaginaries and networked publics: a cultural history of encryption technologies, 1967-2017

PDF

Driving change: copyright, car modding, and the right to repair in the digital age

PDF

Use Signal, use Tor? The political economy of digital rights technology

PDF

Upvoting the news: breaking news aggregation, crowd collaboration, and algorithm-driven attention on reddit.com

PDF

The data dilemma: sensemaking and cultures of research in the media industries

PDF

Virtually human? Negotiation of (non)humanness and agency in the sociotechnical assemblage of virtual influencers

PDF

Political coming-of-age in the era of WeChat: understanding the ethnic media ecosystem and group politics of first-generation Chinese Americans

PDF

Measuing and mitigating exposure bias in online social networks

PDF

The out field: professional sports and the mediation of gay sexualities

PDF

Essays on the unintended consequences of digital platform designs

PDF

Emergent media technologies and the production of new urban spaces

PDF

Labors of love: Black women, cultural production, and the romance genre

PDF

Food is the medium: food movements, social justice and the communication ecology approach

PDF

Essays on digital platforms

PDF

Black Twitter: a cultural, social and political goldmine and why brands should pay close attention to Black audiences

PDF

MATCH CUT: the making of professional screenwriters and a (counter)storytelling movement in film school

PDF

Emotion culture through the lens of PUA (pickup artist): emotions and the social media narratives of structual gendered violence

PDF

Asian American media activism: past, present, and digital futures

PDF

Mobile apps and services for development: What can we learn from the non-smartphone era in ICT4D?

PDF

Temporal dynamics in the lives of health practitioners

Asset Metadata

Creator Corry, Frances (author)

Core Title Sunsetting: platform closure and the construction of digital cultural loss

School Annenberg School for Communication

Degree Doctor of Philosophy

Degree Program Communication

Degree Conferral Date 2022-08

Publication Date 07/22/2024

Defense Date 05/09/2022

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag cultural memory,deletion,digital memory,Friendster,GeoCities,internet studies,memory studies,myspace,OAI-PMH Harvest,platform studies,social media platforms,social networks,sunsetting,Vine

Format application/pdf (imt)

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Dunbar-Hester, Christina (committee chair), Ananny, Mike (committee member), Bowker, Geoffrey (committee member)

Creator Email corry@usc.edu,frances.corry@gmail.com

Permanent Link (DOI) https://doi.org/10.25549/usctheses-oUC111373692

Unique identifier UC111373692

Legacy Identifier etd-CorryFranc-10893

Document Type Dissertation

Format application/pdf (imt)

Rights Corry, Frances

Type texts

Source 20220722-usctheses-batch-959 (batch), University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the author, as the original true and official version of the work, but does not grant the reader permission to use the work if the desired use is covered by copyright. It is the author, as rights holder, who must provide use permission if such use is covered by copyright. The original signature page accompanying the original submission of the work to the USC Libraries is retained by the USC Libraries and a copy of it may be obtained by authorized requesters contacting the repository e-mail address given.

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA

Repository Email cisadmin@lib.usc.edu