ARPANET (1969-2019)

Camille Paloque-Bergès & Valérie Schafer. Internet Histories. Volume 3, Issue 1, March 2019.

“The Internet celebrates its 42nd birthday today, putting it on the cusp of middle age and making its sign an Aries, which is ‘masculine’ and extroverted, says the Internet. But maybe it was actually born in September? Or October? Is it even 42? Some people swear the Internet doesn’t look a day over 28. Here are some the leading claims for days the Internet was born.

April 7, 1969

The first Request for Comment document is drafted by an engineer on the Pentagon’s ARPAnet project, a precursor to the modern Internet. [ …]

September 2, 1969

UCLA computer science professor Leonard Kleinrock establishes the first local connection between two computers in his lab. Matthew Moore of The Telegraph called this ‘the most appropriate’ of all the anniversaries. Discovery magazine and National Geographic both identify this as the true date of birth.”

~ Gustini R. (2011). “Happy Birthday Internet. Today is one of the many dates people cite as the Internet’s birthday.” The Atlantic, 7 April.

“Happy birthday, Internet! You may be turning 45 today, but we swear you don’t look a day over 30. […]

How do we define the invention of the internet? It’s a question that scholars and armchair historians have debated for decades. Did it start with the birth of the web? Did it start with the adoption of TCP/IP? You could make a case for either. But one seminal moment in the creation of the internet cannot be denied: the first host-to-host connection of the ARPANET between UCLA and Stanford on October 29, 1969. At 10:30pm.”

~ Novak M. (2014). “Happy 45th Birthday, Internet!” Paleofuture, 29 October.

The quotations above, among the many which testify to the recurrent online celebration of 1969 as the birth year of the Internet, raise several points. First of all, in both the specialised media (Wired, Gizmodo) and the general press (The Atlantic), ARPANET is celebrated as the ancestor of the Internet, with the emphasis on two milestones in particular: the first connection—the first local connection or the host-to-host connection between UCLA and Stanford—and the first Request for Comments, which came to symbolise the network’s organisational openness. These examples also point to recurrent confusion in the press between ARPANET, the Internet and the Web, as well as to technical but also historical shortcuts, partly linked to the journalistic context and the need to simplify information for the general public. But beyond this simplification, there is also a form of self-celebration of the “network of networks,” a self-referential attitude that Alexandre Serres noted in his thesis on the history of ARPANET back in the early 2000s (Serres, 2000, pp. 29-30):

“We know to what extent the Internet is a comprehensive, self-sufficient ‘medium’, whose ultimate aim has always been found in itself: the network is used above all to communicate through and for the network, and the many technological achievements gain legitimacy on the network itself. This self-referential nature of the network of networks, which has been pointed out by many observers and is one of its major features, also applies to its history, in two ways:

  • By the proliferation, on the network, of websites devoted, in part or entirely, to the evocation of its origins;
  • By the increasing number of accounts by those involved in the development of the Internet themselves, most of whom are still alive and hold positions of responsibility. These players (engineers, professors, administrators, etc.), many of whom have developed their own websites, are cited by many passionate Internet users, who try to collect scraps here and there of the ‘big story’ of the Internet’s origins.

The Internet is thus its own resource for any story to be told […].”

This special issue, which includes a translation of part of the conclusion of Serres’ thesis on the emergence of ARPANET (a work which is relatively unknown in the English-speaking community), two original articles and two interviews, is also linked to the celebration of the 50th anniversary of the ARPANET and the desire to reconsider, 50 years on, the birth of this heterogeneous network that has left such a deep mark on the history of data networks. If we consider ARPANET as a first step towards the Internet, how did it determine—and how does it continue to determine—our current understanding of the Internet? What is left to explore and discover in ARPANET’s history? And if ARPANET is dead today, are its legacy and spirit still alive, and if so, in what way? While the few texts collected are of course not intended to provide all the answers, we hope that they will at least give food for thought.

ARPANET as Part of the Internet’s Official Heritage

ARPANET’s story is part of the Internet’s official heritage, as a first crucial step in its development. Janet Abbate’s seminal work Inventing the Internet (1999) has extensively covered its history and is still a landmark in this area. But the 50th anniversary of ARPANET provides an opportunity to reflect on existing histories and to open up the debate to new perspectives and approaches.

The historiography was first based primarily on accounts from stakeholders and eyewitnesses, keen to build their own heritage of the last decades. The pantheon of historical figures involved in the Internet is dominated by ARPANET protagonists. Using a range of expressive media, from written texts to interviews and even discussions on electronic mailing lists, ARPANET’s memory makers have compiled a solid stock of anecdotal accounts that have had a lasting influence. Even before these oral histories were institutionalised in the 2000s, they were largely relayed, built on and amplified by journalistic narratives throughout the 1990s—the first decade when the digital imagination was unleashed in the public sphere (Flichy, 2007; Turner, 2006), giving rise to a considerable number of “chronicle histories,” as noted by Serres (2000). Their influence can still be felt today. Among them, Hafner and Lyon’s Where Wizards Stay Up Late clearly identifies ARPANET as a central locus of the “origins of the Internet” (1998), focusing on the key role of the US research organisation ARPA (Advanced Research Projects Agency), set up in 1958 and funded by the Department of Defense (DoD). Within this organisation, the Information Processing Techniques Office (IPTO) was seen as the major scientific and political body for the design of ARPANET, with important contributions being made by ARPA contractors. Stakeholder accounts strengthen the heritage of the ARPANET/Internet narrative and keep the memory alive. This is illustrated, for instance, by the “Internet history” mailing list hosted at the Postel Center, which has served to collate an official set of memories and still frequently receives new contributions of stakeholder accounts. ARPANET also remains an important topic, for instance in the IEEE Annals of Computing and Communications of the ACM, which contains papers by stakeholders as well as academics.

In terms of academic research on the topic, ARPANET was first mentioned in studies about global networks in a society marked by new technologies, cultures and economics of information and communication (Castells, 2001; Edwards, 1997; Mattelart, 2000). Its emergence was attributed to the Cold War and to post-World War II technical and scientific innovations in the area of cybernetics. Together with research on the projects hosted by IPTO and ARPA, this has since become a leitmotif running through all accounts of Internet public history.

By the end of the 1990s, research including that by historian Janet Abbate and sociologist Patrice Flichy had paved the way for a more social background, in particular by emphasising the interactions between scientific, industrial and political circles, the various ideologies and spheres of innovation at work among ARPANET’s inventors, but also the increased role of communities themselves in these innovations (Abbate, 1999; Flichy, 2007 (2001)). From the perspective of a sociology of scientific and technical communications applied to ARPANET, Alexandre Serres shed light on the processes of “translations” among these circles, particularly under the influence of cybernetics, and highlighted the absence of a single causality for innovation. Fred Turner also explored these processes from the perspective of media and cultural studies, examining the “mediations” at work (e.g. the movement of ideas from one sphere of practice to another, for example the way in which the “cybernetic metaphor” was circulated from researchers to informed users through the ARPA community (Turner, 2006)). However, by giving centre stage to Stewart Brand, a forerunner of the counter-cultural collectivist (or “communalist”) spirit in digital network innovation and discourse, his book shows other influences on the history of the Internet, both parallel to and interacting with ARPANET. His pioneering work also demonstrates the complexity of innovation trajectories and the absence of a single paternity, more generally recontextualising this innovation and its “processes,” to use a concept of which Alexandre Serres is very fond.

In this literature based on an official heritage, the “original” network not only stood for the origins of the Internet; it also served as a starting point for reflection on plural histories of digital networks. Once again, stakeholders pre-dated academic accounts in their attempts to decentralise the study of these histories, firstly from a geographical point of view: for instance, accounts of the European or Asian development of digital networks have been written in major part by protagonists or eyewitnesses, but always in reference to their US cousin (Carpenter, 2013; Chon, 2013-2016; Davies & Bressan, 2010; Kirstein, 1999; Lehtisalo, 2005; Martin, 2012). This relative decentralisation was accompanied and stimulated by a desire to explore other threads and aspects of technical innovation to understand digital network history. One example is cooperation and competition in the area of standards and protocols (McKenzie, 2011), in which the Internet protocol suite TCP-IP, developed in relation to ARPANET, “competed” with the OSI standard model. First recounted by trade journalists, the protocol angle opened new perspectives on a global history of digital networks while emphasising an international technical pro-Internet culture inherited from the ARPANET community (Malamud, 1992, 1993; Salus, 1995). Since then, these narratives have been enhanced by research on communications and historical studies on innovation policies, with ARPANET continuing to be a central reference, but with a more nuanced international perspective reflecting national and regional circumstances, especially in political terms (Badouard & Schafer, 2014; Russell, 2014; Shahin, 2006). Another example of this dual dynamic of collaboration and rivalry in relation with the ARPA community—symbolised as the archetypal “Netville” (King, Grinter, & Pickering, 1997)—is the role of computing cultures themselves, highlighted by stakeholders, journalists and then scholars. One well-known example of this is the close involvement of Unix user communities and the open computing culture in the global promotion of ARPANET and the Internet (Hauben & Hauben, 1998; Kelty, 2008; Paloque-Berges, 2013; Salus, 1994).

Finally, the tools and materials used to tell the history of ARPANET, such as mailing lists (as mentioned above), have engaged the field of media and communication studies. ARPANET’s communication practices have been qualified as a referential model for Internet history, once again firstly by journalists (e.g. Rheingold, 1993), with an emphasis on users’ power and creativity, which has had a lasting influence on research into digital culture. The fields of computer-mediated communication and cyberculture studies (and later Internet studies), which emerged in the 1990s, have incorporated the model of ARPANET’s mailing lists as a prototype for research about Internet forums and computerised social networks (Jones, 1995, 1998). These communication methods and their role in building Internet culture, memory and heritage from ARPANET to the Web have already been partly highlighted (Paloque-Berges, 2018).

“Re-inventing” ARPANET?

When it comes to opening new perspectives on the history of the Internet in general, recent journal issues have paved the way (Brugger, Goggin, Milligan, & Schafer, 2017; Haigh, Russell, & Dutton, 2015). In terms of the history of ARPANET, at least three new avenues of research are expected to shift established perspectives, and the creativity of historians and research will undoubtedly continue to open up further avenues in the coming years and generations.

The first of these, already anticipated by Alexandre Serres at the end of his thesis in the early 2000s, is well under way, shifting the focus from US-based histories of the Internet and ARPANET and allowing us to consider developments from the perspective of other world regions. This change in perspective was brought about in particular with Russell’s book (2014), mentioned above, which looks at the largely European history of the OSI model, and Schafer’s thesis (2012) on the French Cyclades and Transpac networks. Some of the protagonists in these stories have strong links to ARPANET, such as Hubert Zimmermann, who contributed to the design of OSI, the great rival of TCP-IP during the protocol war, and Louis Pouzin, whose datagrams were mentioned by Vinton Cerf as a source of inspiration for the Internet.

Also in 2017, the first Routledge Companion to Global Internet Histories (Goggin & McLelland, 2017) and the international conference on “Computer Networks Histories” showed how much ARPANET influenced the spread of digital networks throughout the world, especially through academic and industrial channels (not only in Europe but also in Brazil and India), while underlining the complex constraints at different geographical and geopolitical scales that made these histories unique. It also shed light on computer network histories that were not, or at least not directly, influenced by ARPANET, for example in Russia and China (on the Soviet Union, see Peters (2016)). The internationalisation of ARPANET’s history is concomitant with critical perspectives on teleological readings of the innovation trajectories of digital networks, as underlined by Christopher Leslie (in Tatnall & Leslie, 2016).

A second avenue draws attention to players in ARPANET’s history that did not tend to occupy centre stage, despite their significant involvement. Network computers, starting with ARPANET, are still historically documented extensively from the upper levels of digital interfaces to the lower levels of infrastructures, by way of protocols, data transfer and other software and hardware issues. This involves materials and practices that were produced not only by creators and administrators, inventors and managers, but also by programmers, developers and users, even amateurs, as well concurrent or allied communities. The re-invention of the historiography around ARPANET involves looking into documents and players that have attracted less attention over the years. Some aspects had remained invisible, for instance due to their technical features, deemed less important or less accessible. One significant effort made in this direction is the research by Bradley Fidler and Morgan Currie into the ARPANET maps, which has revealed issues of aesthetics, design and representation in the very crafting of infrastructures and shed light on an invisible topography in the history of the Internet (Fidler & Currie, 2015, 2016). Also worth mentioning is the very recent and important article by Fidler and Russell (2018) which focuses on ARPANET’s infrastructure. The topic of cryptography is also related to this investigation of the less visible parts of ARPANET’s history (DuPont & Fidler, 2016).

Another task, which is still ongoing, is to examine which research centres made use of ARPANET and to explore the roles, practices and ambitions that have not yet been fully analysed—in short, a focus on users. Abbate (1999) already called for emphasis to be placed on the crucial role of users (with different skills and involvements) in the constant re-invention of the Internet. Early user communities at the fringes of ARPANET have seldom been a focal point in sociological and anthropological research (e.g. Kelty, 2008; King, Grinter, & Pickering, 1997), although the theme has re-emerged in recent historical investigations (Paloque-Berges, 2017; Tatnall & Leslie, 2016). Stephanie Dick’s presentation at SHOT St Louis 2018 on the MIT Macsyma (Mac Symbolic Manipulator) project is also worth mentioning in this respect: she pointed out that the PDP 10 was one of the most popular nodes in 1972 within ARPANET but also that it was very difficult to become a user of the Macsyma system; as she explained, the community formed around the Macsyma project was not pre-existing but developed as the project progressed.

A third avenue, in a similar vein to research on less-considered stakeholders, takes us outside the strict scientific, industrial and technical sphere of ARPANET and looks at broader cultural and social matters: research involving economics, demographics and social or environmental issues. Economic questions have been raised in analyses of the debates on business models that took place during the development of ARPANET in the 1970s (Day, 2016), in line with the tradition of business history that is largely ingrained in the history of science and technology in general and computing in particular. But to our knowledge, other social issues have not been well researched in the context of ARPANET, even though they are indeed part of the new agenda in the history of computing. This could be explained by historiographical exclusion. Indeed, computer scientist Lynn Conway (2018) has cited ARPANET as an example of what was recently called the “Conway effect,” namely the disappearance of minorities from the history of science and technology. Gender does nevertheless constitute a small exception, although not a major one, since ARPANET still remains largely absent from the main contributions to gender-conscious histories of computing and networks (Abbate, 2012; Ensmenger, 2012; Misa, 2010; Schafer & Thierry, 2015). It is telling that only very few women have been celebrated in ARPANET’s heritage (Evans, 2018). Examples include Joyce Reynolds, who was closely involved and worked with Jon Postel on the early development of IANA, and Elizabeth “Jake” Feinler, who took care of the network’s early documentation and remains a prominent figure in maintaining documentary heritage (Feinler, 2010; Schafer, 2010). Indeed, ARPANET’s “Netville” is still under the spell of “IMP guys,” the white, male scientists and engineers who created and operated the interface machine processors that were at the centre of the innovations related to the Internet (Waldrop, 2001). However, ARPANET itself hosts unheroic figures in its shadows, and focusing on them seems crucial today in order to produce critical analysis and alternative or counter-narratives in the trajectories of digital networks, from victorious innovations to failures by way of the many “missing net histories” (Driscoll & Paloque-Berges, 2017). Recent calls to highlight the role of under-considered players are increasing, as seen, for example, in the “maintainers” manifesto, which aims to foster research about everyday workers who are hidden behind the screen of the more “glorious” innovators in the history of science and technology (Russell & Vinsel, 2016). One exciting contribution of such an approach is provided by Fidler and Russell (2018). Their introduction emphasizes their desire to break with traditional historiography, which can focus too narrowly on innovation:

We propose a methodological and ontological reversal in the history of networking. We aim to prioritize infrastructure and its maintenance as the ontological primitive to be explained—and to leave innovation aside as a taken-for-granted process for which explanation is unnecessary. As such, we will trace out the first steps at a history of the maintenance of network infrastructure, using the archetypical case of the Arpanet and its place in the early Internet as our case study (Fidler & Russell, 2018, p. 901).

This Issue

The aim of our call for papers was to revisit the history of ARPANET, its genesis, development, heritage, memories and the writing of its history 50 years after the first four nodes came into service. We received several proposals that set out to summarise ARPANET’s rich historiography without really being in a position to shed new light on it. This observation is in itself reassuring for the fruitful historiographical work carried out in the early stages, and it hints at the presumed rarity of documents and new facts still to be discovered. It also gave us the opportunity to introduce Alexandre Serres’ work, which was completed in 2000, to a non-French-speaking audience (as we had always intended to do), since some of his conclusions remain relevant or are echoed in the articles and interviews in this special issue.

Although in a 2015 issue of Information and Culture on the history of the Internet, William Dutton, Thomas Haigh and Andrew Russell were able to write –

Our perspectives on its history have remained focused on the men who originated it and, as the ‘greybeards’ of the Internet Engineering Task Force, continue to provide technical leadership in the evolution of its protocols.

The Internet which is invented in Abbate’s book, and the other works mentioned, is the Internet as understood circa 1994, not the incomparably broader Internet of 2014. (Haigh et al., 2015)

—it should at the very least be stressed that the history of ARPANET seems less subject to renewal, or even to controversy. However, some specific parts of its history are debatable, such as the origin of email, which has been the subject of particularly fierce paternity disputes.

We can certainly still find discussions, particularly on packet switching and the respective roles of Paul Baran, Leonard Kleinrock and Donald Davies, enlightened by Morten Bay, while emphasising that it is necessary to accept complexity and a Zeitgeist that goes beyond a debate on single paternity, and especially to understand “the sequence of events that led to its inclusion in the ARPANET design.”

While Morten Bay uses a few previously unpublished documents, the major sources seem to have been widely identified and accounts from pioneers are already numerous, gathered in particular by the Charles Babbage Institute at an early stage. This in no way reduces the value of the oral interviews published in this issue, whether that of Larry Roberts by Morten Bay, which is part of a series following on from those published in previous issues with Paul Baran and Leonard Kleinrock, or that with Michel Elie and Gerard Le Lann, which tells the story of how two French protagonists participated in the history of ARPANET. Although we have already heard these protagonists’ accounts, we may come to discover their voices, their unique personalities, beyond their factual contribution to the collective project. Far from trivial, this sheds light on how the stories about ARPANET were and are still told (and by whom, and who is still approached and listened to) and thus transmitted, helping us to critically understand the heritagisation and historiography of ARPANET. For instance, it is especially interesting to compare the individual perspective (and personal quarrels) in Robert’s testimony with the insistence on the contribution of the collective and the emphasis on scientific culture in Le Lann’s, or the reference to a social and cultural context beyond academia in Elie’s.

Finally, Fenwick McKelvey and Kevin Driscoll offer a new perspective by integrating Science and Technology Studies and a reading in terms of boundary objects. Considering IMPs, which had already been the subject of an article by David Walden in 2014 in the IEEE Annals of Computing, they propose here a new approach that also draws parallels with the emergence of digital studies, platform studies and code studies and explores the “trading zone,” to refer to the notion coined by Galison, that ARPANET may have created, enabling it to be seen in a conceptually different way. It should be recalled that the notion of “boundary object” owes a lot to Star and Griesemer (1989), who provide an analytical description of those processes in which actors from different social worlds manage to cooperate despite their diverging points of view. This notion is, indeed, to be handled with care—as Susan Leigh Star herself suggested: since its creation, the concept has mostly been used for its “interpretive flexibility,” i.e. a property that allows it to support heterogeneous translations and different types of knowledge. However, as Trompette and Vinck (2009) emphasise, other dimensions have been overlooked; for example, boundary objects incorporate sets of conventions, standards and norms that are typical of specific communities of practice; they allow us to account for the processes of delegation of work or other activities or for the performative action of artefacts in the production of knowledge (Musiani & Schafer, 2018). It is the combination of these negotiations, translations and sets of conventions that the article “ARPANET and its boundary devices: Modems, IMPs and the inter-structuralism of infrastructures” fully highlights:

[…] the IMP as historical boundary object that helps expose ARPANET’s close relationship to the telephone system. Our analysis offers a novel history of ARPANET as a repurposing of the existing telephone infrastructure. Beyond the historical contribution, this approach has wider implications for the theory of media infrastructures, specifically the ‘interstructuralism’ of ARPANET and the nature of borders between seemingly disparate social, political, and technological regimes. (McKelvey & Driscoll, 2018 in this issue)

Their study also reintroduces a sometimes-neglected element, telephone heritage, which also makes it possible to connect the history of ARPANET with that of telecommunications, media, even transmedia and convergence, in a narrative that previously often tended to emphasise new developments rather than continuity. How do we move from one network to another, from one community to another, from one period to another? How are they linked? How do we study socio-technical correlations beyond mythological narratives? The common carriage model and heritage, the convergence of sociotechnical systems and IMPs as a mediating technology between AT&T and ARPANET are examined from a new perspective, allowing us to consider the IMP as both gateway and translator.

It is worth noting that in his thesis on ARPANET, Alexandre Serres had already applied concepts drawn from the sociology of translation to RFCs, together with the actor-network theory, which has strongly influenced STS.

The interest in RFCs shown in particular by the quotations that open this introduction but also acknowledged in Michel Elie’s account invite us to consider their central role; the date of the first RFC is still considered as highly symbolic. Paul Baran’s diagram on the centralised, distributed and decentralised vision of a network continues to be a reference. While Alexandre Serres mentioned the recurrences of “ARPANET þ history” on AltaVista, it is also worth noting the similar test carried out by Paul Baran (an image search in Google gives millions of results). But the mention of RFCs is also recurrent, this time reflecting a distributed and peer-to-peer organisational framework. Sandra Braman’s work (2011, 2012, 2013) has provided fruitful analyses of RFCs, which today remain the symbol of a pioneering spirit open to contribution and peer-to-peer methods, as Michel Elie remembers when he recalls his time at UCLA in 1969-1970 and the contributions he made.

In addition to the fact that RFCs constitute an extraordinary memory of the network for historians, this example of openness based on technical competence, informal and peer-to-peer discussion more flexible than traditional standards bodies, would foster both the adaptability of the network and also a process of broad stakeholder involvement.

This phenomenon of creating interest through openness and participation is not negligible. Openness guaranteed publicity. As Vinton Cerf (1990, pp. 22-23) noted:

A guy named Gerard Le Lann (sic) was at IRIA working with Pouzin and came to my lab at Stanford for a year and had a lot to do with the early discussions of what the TCP would look like. So did Bob Metcalfe, it turns out. Metcalfe was at Xerox at the time and in June of 1973 we began working together, Le Lann, Metcalfe, and I, on the design of the host-to-host protocol for INTERNET. Eventually Metcalfe got impatient with the rate at which things were going. I was trying to get a large number of people to agree on a set of protocols, and every time you brought in a new player we had to go through the argument again. Meanwhile Metcalfe had five or six guys over at Xerox trying to get the local area nets running. Finally they said they didn’t want to wait until this process of agreement and consensus finally concluded, so they went off on a slightly different tack and invented XNS that took some different choices than the TCP did. And they got it up and running before ours, in fact. Of course in the long run we’ve… They kept it secret, and that was a mistake, I guess, now looking back. If they hadn’t kept it secret, we might all be using XNS instead of TCP. But as it stood, TCP turned out to be the open protocol that everybody had a finger in at one time or another. That is just how it all worked out.

This symbol of ARPANET’s intellectual and organisational openness has become an aspect that is regularly cited in the Internet neutrality debates or by those who criticise the excessively unilateral governance that currently characterises the Internet and the confiscation of the spirit of openness by companies or countries.

But it is also worth considering, beyond this symbolic heritage, a technical heritage that Gerard Le Lann emphasises when he refers to smart vehicle or energy networks in which distributed network pioneers like Robert Metcalfe are involved. This suggests that we need to think not only about “When old technologies were new,” to quote Carolyn Marvin (1990), but also when new technologies are old.