Benjamin Peters
The map is not the territory.
—Polish American philosopher Alfred Korzybski
Computer networks cannot save us—nor can we save them. Often, network designers and users do not even know what networks are: a network is not what we think of as a network. A network is not its map.
“It’s not that we didn’t think about security,” admitted internet pioneer David D. Clark to the Washington Post about the early days of the internet: “We knew that there were untrustworthy people out there, and we thought we could exclude them.”1 In the fifty years since 1969, not only has the internet not excluded bad actors, it has grown from connecting a few dozen scientists to encompassing over four billion users, more than half the world’s population. Hundreds of millions more are likely to go online in Africa and Asia in subsequent years; ambitious predictions even hold that almost every living person and far more things (flip-flops?) will be online by 2030. The internet is an unparalleled access machine. But towering successes cast long shadows: what is this unparalleled growth for? Access to what and for whom? Whom does it all serve? It should not surprise us that no other class of actor has benefited more from the enormous global reach, speed, and capacity of the internet than the organizations that prop it up. Not the users. Not the visionary network designers. Other networked organizations—from the undersea cable owners to service providers to big data brokers to data aggregators—have gained the most. While surely not all organizations intend to exploit people, it remains the historical rule that at least a few will, and in so prospering just a few bad actors may normalize exploitative behavior that serves networked organizations in general: networks now let us exchange data but at the cost of rendering ourselves products on the network subject to network surveillance, discrimination, gatekeeping, platforms rigged to reward addictive behavior and groupthink, toxic attention economies, and much else. Looking back, one cannot help but counterfactually wish that a more secure design at the dawn of computer networking might have paid dividends now, half a century on.
But that would be sloppy thinking, for it is precisely such design-oriented solutionism about networks that led to the current state of computer networks in the first place. It is shortsighted analysis (that often imagines itself to be structural, fundamental, and far-reaching) that justifies a network based on the values embedded in its original design. Why are the values we embed in the design of a network not enough? It is simple: a network is not what we think it looks like. Networks may enjoy attractive visual design-reality homeomorphisms, but they do not mean what we think they mean. A live complex computer network—its uses, its meaning, its consequences—is not what it looks like on paper. A network design is defined visually and spatially through architectures, assemblies, gaps, links, maps, nodes, points, topologies (such as centralized spoke-and-hub, decentralized, distributed mesh, free scale, daisy chain), etc. By contrast, the internet in situ is more than the sum of its complex uses and users: thousands of thousands of thousands of users shuffling data between organizations for as many reasons—or no reason at all—at any given time.
How should we make sense of the internet then? Philosophers have long bristled at the fact that technology is often understood only instrumentally—by its use value. Yet network analysis often does not even get that far. Consider the analogy to money. Money and networks are both immensely useful, but by reducing their value into a single price or an ideal design, analysts multiply their potential abuses. When it comes to money, few ignore the difference between what money can be exchanged for (products for a price) and what it can be used for (products for use). By contrast, for fifty years, popular networks have extrapolated from what a network looks like to what it can be used for. Observers scowl at centralized network schematics (one hub with all spokes), judging them good for nothing but maybe centralized control and surveillance situations like a panopticon. Distributed networks (each node connects to its neighbor nodes), meanwhile, appear to share liberal democracy’s commitment to all nodes being equal, locally connected, and robustly integrated. Given those two ideological endpoints, how, exactly, did the world’s most distributed network end up a panopticon?
Modern-day design analogies go wrong not because network visualization tools—graphs, dynamic models, schematics, etc.—are not yet good enough to represent network nodes and their gaps (even if the roughly fifty billion nodes on the internet stretch even machine abstraction). Rather, it is that such network design visuals are themselves the gaps in the modern understanding of networks.2 To twist that old line often attributed to the mathematician and early information age polymath Norbert Wiener, the best model of a cat is a cat, preferably the same cat: so too is the most reliable model of a complex large-scale network that same network at work in the world. A network on paper or screen is no computer network, and often computer network models obscure what embedded organizations might actually use the network for.
No historian will be surprised by the assertion that technology does not behave as it is initially designed: it is common sense that technologies often develop unforeseen affordances. For example, when the telephone was introduced into the US in the late nineteenth century, it was sold to middle-class businessmen with the promise of globalizing their companies; however, once they adopted it, women turned to the telephone in greater numbers than businessmen to connect with local friends and contacts. Only a fool clings to the stories we tell ourselves about new media. No one today insists that telephones are inherently global business tools: telephone sets have largely vanished, although I’m sure we could arrange a call to discuss it more. By analogy, there is no reason to imagine networks as fundamentally democratic, and yet we do all the time.
Network designs come with a second-order recognition problem that stems from the fact that all sufficiently large networks are necessarily the products of multiparty collaborations. (It takes a lot of groups to make a network.) Therefore, it is impossible to view a network from any single perspective. Instead, we insist on seeing too much in their network designs. By believing that we embed values into the universal design of networks, we wrongly attribute explanatory power to such design values.
An analogy obtains in corporate social responsibility. As media and organization scholar Siva Vaidhyanathan points out, blue-chip corporations have come to embed lofty values directly into their mission statement and operations in order to make a positive difference in the world while also gaining internal and public buy-in for their brand. No matter how genuine in intent, the more responsible and desirable these values appear, the more corporations or networks can counterintuitively deflect, defer, and delay the normal correcting hand of history—the checks and balances of industry self-criticism and state regulation meant to ensure socially responsible behavior. Everyone should want and demand more socially responsible network corporations in practice, of course. But, for the same reason that my planning to be a more caring neighbor does not make me so, designing values into our corporations or networks offers no guarantee they will behave accordingly.
This essay critically considers and compares early attempts to design large-scale computer networks, including the internet. In the 1960s and 1970s, engineers designed distributed packet-switching computer networks after the American values at the time of surviving a nuclear war, modeling democracy, market institutions, and collective smarts. Network companies continue to fashion themselves after the image of the society they reflect: perhaps no other company does this more than Facebook, a self-professed “community” of nearly two and a half billion active users that promises (like much in your cable company mailer) to increase your contact with loved ones, float you in an ocean of entertainment, amplify your voice on forums, and flatten your access to goods and services. And yet the same network companies have shaped the experience of the internet in ways their designers did not intend for reasons they are not trained to identify. Of course, not all network companies are malicious masterminds with mind-control rays hell-bent on speeding society suicide; most successful ones are, by all accounts, disciplined to well-intentioned design values that invite both our celebration as well as vigilant checks. For example, network neutrality policies are noble and vitally needed, but nothing about network neutrality should be mistaken as actually being neutral: in resonance with Sarah Roberts’s chapter in this volume, the public claims about design values—the forward-thinking solutionism of artificial intelligence (AI) and machine learning is another one—obscure, not disclose, the causes and effects of actual network behavior.
A proposition: no one today knows exactly how or why actually existing, large-scale networks behave in the ways they do, and in the meanwhile, surveillance has become the quiet coin of the realm: the logics of surveillance, by contrast, are well understood. Surveillance online today is not some benign anonymized monitoring that feeds and informs public understanding about our own behavior on networks. No, our networks exist in an age when surveillance services compete to buy and sell information about you, without your wishing it, and every other “you” to the highest bidder. Such “smart” surveillance has made idiots out of everyone and everything that plugs in in the original Greek sense of private persons: it privatized every action. Online we are not even persons—we are personas. Forget Russian bots: you are the internet’s favorite bot.
Critical interventions into our networked worlds, such as this and Kavita Philip’s essay in this volume, are sure to prove both necessary and insufficient: this essay calls attention to the many variables that have shaped our global network history without informing network designs or discourse; in particular, it attends to how institutional behavior has shaped large-scale networks in ways that quietly undermine their designs’ publicly celebrated values. No one can know yet what a network is or does. This yawning gap—strung out here in an attempt to compare the history of three national computer network projects—remains pressingly open.
Nothing less than the future of networks hangs in the balance.
The remainder of this essay compares the stated design values and actual institutional practices across the histories of three national computer network projects since 1959: the Soviet OGAS project, the Chilean Project Cybersyn, and the American ARPANET. First, on the eastern side of the iron curtain at the height of the Cold War tech race, Soviets were designing major computer network projects dedicated to explicitly civilian purposes: namely, Soviet scientists and statesmen sought to upgrade the entire paper-based command economy to the speed of electronic socialism, but what happened next simply does not follow the high-noon scripts for Cold War showdowns. The complex institutional conditions in which these Soviet network projects tried to take root tell a different story that scrambles and refreshes mainstream network thinking since the Cold War.3 Second, as eloquently chronicled in Eden Medina’s pathbreaking book Cybernetic Revolutionaries, the relevance of broader political, organizational, and cultural variables becomes even more obvious in another major socialist cybernetic national network. Namely, the Chilean President Salvador Allende’s computer network project, Project Cybersyn, sought to democratize and socialize by computer network parts of the Chilean national economy between 1971 and 1973. In the process, the network was designed to hand the reins of control over to worker-citizens and Allende’s statesmen.4 Here, too, the historical arc of Project Cybersyn reminds us how poorly design thinking predicts realities. Our third vignette, on the US ARPANET, likewise bears none of the neural or military design imprints that inspired it then or now: instead, the ARPANET, which went online in 1969, sped scientific exchange and, once paired with the personal computer revolution, grew into the contemporary internet.5 (If pressed to speculate, I would bet that future historians will remember the globalization of the internet primarily for its uneven consolidation of knowledge and power among populations that write in world empire languages and code, but no doubt the future will unsettle such predictions at least as much as does reflection on the past!) These historical surprises have generated countless new innovations as well as cooptations, including some of the internet’s most celebrated design values—such as antihierarchical packet-switching and democratizing distribution designs. This essay turns against visual network-design values and toward the explanatory weight that complex institutional behavior has had and will continue to have in shaping a less straightforward, more critical, and more incendiary history of how computer networks came to encircle and then enchain the world.
In the fall of 1959, the young star military researcher Anatoly Kitov wrote General Secretary Nikita Khrushchev with an apparently original thought: why not connect the nation’s military computers into one giant network that could benefit civilian problems, like economic planning? It was the height of the Cold War technology race, and the Soviets had all the mathematics, motivation, and military might to build the contemporary of the Semi-Automatic Ground Environment project in the 1950s or the ARPANET in the 1960s. At that very moment, Soviet scientists were expanding nuclear power, launching satellites and spacecraft into orbit, and making huge advances across the theoretical sciences, including innovations in an indigenous computer industry into the mids1960s. Why, then—after thirty years of attempts—was there never a “Soviet internet”?6
Between 1959 and 1989, a group of leading Soviet scientists and statesmen and women agreed there should be a national computer network to manage and digitize the nation’s paper-based command economy, thus ushering in an era of electronic socialism. By circumventing all the cumbersome delays in paperwork and subsequent perpetual backlogs in the supply chain, every factory worker and economic planner with access to the proposed national network would be able to interact, exchange reports, and make real-time adjustments to their contribution to the command economy—all at the sublime speed of electricity. The result, they hoped, would be no less than what Francis Spufford has called, in the title of his delightful historical novel, the age of Red Plenty.7
Under the direction of Soviet cyberneticist Viktor M. Glushkov, the All-State Automated System of Management (obshchee-gosudarstvennaya avtomitizirovannaya sistema upravleniya)—or the OGAS project—sought to fulfill such a socioeconomic vision while sharing few of the design values that internet enthusiasts would recognize today: no packet-switching protocols, no distributed design, no peer-to-peer populism. Rather, in order to solve time-sharing and interoperability functions while appearing politically viable to the Soviet state, the OGAS network design was modeled directly after the three-tiered hierarchical structure of the planned economy. As the secretary of the institute in charge of the optimal economic models or software for managing the networked economy once put it, the decision was made to “build the country’s unified net hierarchically—just as the economy was planned in those days.”8 Thus, at the top, the network would link one central computing center in Moscow to a second tier with as many as 200 computer centers in prominent city centers, which would in turn connect to a third tier with as many as 20,000 computer terminals distributed throughout the industrial base of the Eurasian continent. In order to pass as a politically feasible state hero project, the network was designed to be neither centralized nor distributed. It should be, like the state, a decentralized hierarchy, meaning that any user would be able to connect directly with any other user on the network, provided that Moscow had first granted clearance to do so. Worker feedback, criticism, suggestions, and updates could be submitted directly from the factory floor to anyone deemed relevant by a supervisor. All this activity would also, in turn, conveniently permit the supervising state to collect tens of thousands of dossiers on individuals and organizations in the national network.9
Like the Chilean and American network designers discussed here, Glushkov and his team were inspired by the grandiose cybernetic design analogs between information systems in vogue then (as well as today, although the odd label “cybernetics” has been shed in favor of new ones): Glushkov, whom the New York Times once called “the King of Soviet cybernetics,” imagined the OGAS project to be even more corporeal than “brainlike” (mozgopodobnyi). He saw networks as fully embodying the material economy and thus the nation: his national network was designed to function as the nation’s informational nervous system for electrifying and animating the sluggish print-based organic body of the national economy.10 Cyberneticists and design theorists such as Glushkov, Anatoly Kitov, Nikolai Fedorenko, and many others did not make the design mistake popular in the West of imagining information to be immaterial or virtual (see Nathan Ensmenger’s essay in this volume on the Cloud in cloud computing). Instead, they planned the state economic network as if it were a living machine fully embodied according to (yet another visually obvious) cybernetic design value: if the Soviet economy was the national body of workers, it needed not just the state to be its central information processor and brain. Both the body (the nation of worker) and the brain (Soviet state) needed a live nervous system—a real-time national computer network.11
It was a beautiful design in principle. But this is the stubborn point: none of these stated designs—socialist economic reform, decentralized hierarchical states, mass state surveillance, or cyber-network corporeal collectivism—had any bearing on the outcome of the OGAS and other early attempts to build Soviet networks. Unanticipated obstacles interfered: even though Glushkov often had the ear of the top brass in the Soviet state (once almost winning over the Politburo to the project in 1970, a handful of years after the project had been under internal review and one year after the ARPANET had gone online), very few others in the Soviet economic planning bureaucracy wanted a computer network’s help in reforming the economy. That was their job, not a computer’s, the general sentiment went. Despite the internal boasts and the complaints from capitalist critics, the heart of the Soviet economic bureaucracy was in practice no hierarchy. So middle-level bureaucratic resistance mattered in ways that no top-down decision could every control for. The number of institutions—and bureaucratic ministries—that would have to agree to implement the OGAS project underwent such a continuous “treadmill of reforms” that the resulting traffic jam of competing, conflicted, and self-interested ministers proved impossible to navigate well through the 1980s, when the economy was teetering toward internal collapse.12 For example, immediately after Stalin’s death in 1953, the number of ministries were consolidated to fewer than a dozen; a decade later, after the uneasy decentralization of Khrushchev, the relevant economic ministries would balloon to over thirty-five, with a continuous and uneven pendulum swing back and forth thereafter.13 The resulting institutional ambiguities benefited status quo–inclined economic planners with enough moxie and strategic self-interest to hold onto their positions in such tumultuous internal reforms. The resulting mismatch between stated design values and the actual practices proved enough to shipwreck many efforts to reform the economy, including the OGAS project, against the rocks of competitive self-interest. Informal resistance reverberated up and down the chain of command: the occasional worker would sabotage prototype machines, fearing their own replacement, while ministers in the Politburo, the second highest state committee, apparently mutinied against Prime Minister Aleksei Kosygin rather than cooperate on the network.14 Not only was the cybernetic analogy of the nation as the economic body misplaced, it could not materialize in the unfettered and turbulent interinstitutional competition over power and resources.
This Soviet network story amounts to the signal contradiction that my most recent book begins with: the US ARPANET took shape thanks to collaborative research environments and state funding, while the Soviet attempts to build civilian computer networks were stalemated by its own unregulated competition and internal conflicts of self-interest. The first global computer network took shape, in other words, thanks to capitalists behaving like cooperative socialists, not socialists behaving like competitive capitalists.15
At the same time that the Soviets were stumbling to fund the OGAS project, an uneasy political regime was taking place in Latin America, and with it a fascinating case study of another significant socialist network project—the Chilean Project Cybersyn, designed by the British cyberneticist Stafford Beer under Salvador Allende between 1971 and 1973, now rigorously told in Eden Medina’s signal history Cybernetic Revolutionaries: Technology and Politics in Allende’s Chile.16 In her account, the young engineer Fernando Flores reached out to Stafford Beer in 1971 for help building and managing the 150 enterprises the new Allende regime had under its direction, including twelve of the twenty largest Chilean companies. Together, Beer and Flores design a nationwide computer network that would link together through many feedback loops a single “viable system” of government officials, industry managers, and worker-citizens.17 Like the Soviet case, a networked democratic socialism promised to serve its citizens’ economic needs with additional information, processing power, and economic predictions. Even more explicitly than in the Soviet model, in the Chilean project Beer designed the state as the head of the economic plan, with many sensory inputs making it sensitive to the changing needs of the proletariat classes; unlike the Soviet case (which was developed almost a half century after the Russian Revolution), Project Cybersyn was steeped in the revolutionary design values of a fresh regime change, when internal hopes ran high that technocratic solutions could assuage and ease brewing class tensions. That hope proved short-lived: after surviving a trucker strike, the Allende regime—and Project Cybersyn with it—collapsed on September 11, 1973, after a revolt of middle-class businessmen and a CIA-backed military coup led by Augusto Pinochet.18
Here, too, Beer’s stated design values of democratic socialism and cybernetic viable systems tell us much, although (again) nothing that either the designers anticipated or that correlates with the network’s actual development. Consider, for example, the design choices informing the futuristic Cybersyn “operations room” at the center of the state national network of telex machines. The room held “seven chairs arranged in a circle within a hexagonal room,” deliberately introducing an odd number to prevent voting ties, with walls lined with screens that displayed brightly colored graphic designs signaling economic data and warnings to state officials.19 On the armrest of each chair, a series of large, legible buttons awaited the command of both state officials and citizen-workers—although these designs reveal something about the gendered future of their networked Chile: these “large geometric buttons” were “big hand” buttons designed to be “thumped” when emphasizing a point.20 However, nowhere in the room, in Medina’s careful reading, does one encounter a keyboard, the gateway to the female secretarial classes at the time (and to incisive analyses by Mullaney and Stanton in this volume). The future was designed for male workers and officials who do not know how to type. Introducing a keyboard would “insinuate a girl between themselves and the machinery,” Beer claimed, “when it is vital that the occupants interact directly with the machine, and each other.”21 As Medina holds, “Such design decisions were not neutral. They reflected who the design team believed would hold power in Chile’s revolutionary context and enforced that vision. Male factory workers and government bureaucrats would have decision-making power. Other kinds of workers, such as clerical workers, and women, would not.”22
The historical design values—democratic socialism, socioeconomic justice, and male decision-making classes—reveal volumes about the time and place in which technologies merge with and emerged in society, including a sensitivity to the pressing political and social class issues of the day. However—and here is my critical point—these design values tell us very little about the actual conditions, causes, and complications shaping the network itself. Unlike the Soviet and American case studies, the explicit network design values in the Chile were oriented toward managing the very historical forces that undid it—oriented, that is, in vain: Project Cybersyn addressed but, facing broader political and military opposition, ultimately could not resolve the conditions of class unrest and revolt that toppled the Allende regime.
The Chilean case offers a vital reminder that network technologies, even when oriented toward relevant social values, are not enough to single-handedly change the values at work in society, for (of course) no technology alone can “fix” society since no technology ever contains society, no matter how smart we may imagine its social values to be. The point is not simply that technologies are political (imagine claiming otherwise in the wake of the Russian hacking of the 2016 US elections): it is that network projects are twice political for how they, first, surprise and betray their designers and, second, require actual institution building and collaborative realities far richer than any design—even that of the most self-reflexive cyberneticist—can account for. As the Chilean case shows, every network history begins with a history of the wider world. Once network observers set aside our visual design-value biases, the possibilities previously unseen behind this realization should delight, not disappoint.
The United States government is a prime example of a complex organization. While intervening against socialist regimes in Chile and the Soviet Union, it also aggressively subsidized the cross-institutional collaborations necessary to jump-start the ARPANET, the first packet-switching distributed network that would, in time, seed the contemporary internet. Except perhaps scientific exchange, none of the stated motivations for key ARPANET innovations—national defense, commerce, and communication—have been fully realized in it or its network descendants. Optimistically, this may mean the underlying purpose of the internet is still wide open—and often generatively so. More realistically, this also means the internet remains open to cooptation by increasingly large organizations (corporations and states alike) keen to seize and wield private control over our network relations. Perhaps we can say that if the Soviets never agreed on how to build the OGAS network, no one has yet agreed on why the US built the internet—and there is much to learn from both disagreements.23 The contradictory design values of the mixed American economy are so great—well-managed state subsidy coinciding with anti-socialist, anti-statist ideologies—that we can now see in retrospect that the pathway leading to the internet is marked by historical accidents.24 The internet has never been what we think it is. It will continue to surprise.
As it long has. A few of those surprise twists and turns follow: the ARPANET went online for the first time on October 29, 1969, before growing into a network of other networks for scientific exchange such as the NSFNET, CSNET, and EUNET, even as the main motivation and financial support for several key ARPANET innovations came from the military.25 Indeed, the communication tool most heralded as flattening global communication inequalities emerged out of the paragon example of command-and-control hierarchy: the US Department of Defense. The Advanced Research Project Agency (ARPA), a special research arm of the Department of Defense, was funded in February 1958, five months after the launch of Sputnik, with the initial goal of militarizing space, although soon thereafter its space mission was reassigned to its contemporary civilian agency, the National Aeronautics and Space Agency (NASA). The ARPA instead focused on basic computer research and information systems fit for the age of nuclear-tipped Sputniks.26 Even as its missions molted, however, its institutional funding paths continued to straddle military defense spending and essential nonmilitary basic research funding—at least in the 1960s, before the Vietnam War pressured Congress to limit ARPA spending to specifically military projects, driving a brain drain to Xerox PARC, among other private-sector companies.27 In the 1960s, many network innovations—distributed network design, packet-switching protocols, time-sharing, interface message processor—came together into one computer network. This period, which was perhaps the most generative in US history, stands out for the significant cross-agency collaboration supported by virtually socialist-levels of state funding for nonmilitary basic scientific research. Well-funded collaboration sped the birth of national computer networking.
For example, after some bumps, computer network engineer Paul Baran’s packet-switching distributed networks grew out of the dashes in the military-industrial-academic complex. In 1960, Baran, then a researcher at RAND—a private think tank spun off from the US Air Force in 1956—was tasked with designing a “survivable” communication network: “there was a clear but not formally stated understanding,” he remarked later, “that a survivable communication is needed to stop, as well as to help avoid, a war.”28 He had the strained strategies of mutually assured destruction in mind: in order to maintain peace, the nuclear superpowers would need to enter a perpetual standoff with one another, legitimately armed for the apocalypse at all times. To make good on such a threat, both superpowers would need long-distance communication network infrastructure that could withstand multiple nuclear blasts.
Military survival logics combined to inform Baran’s network design values with another source of unlikely inspiration: the self-repairing human brain in the work of the founding cyberneticist and neurophysiologist Warren McCulloch.29 “McCulloch in particular inspired me,” Baran said, continuing, “He described how he could excise a part of the brain, and the function in that part would move over to another part.”30 In other words, for Baran (like Glushkov and Beer), the biological brain and the social network could become mirror models of survival and adaptation: just as a damaged brain can rewrite itself, so too could a message be rerouted in real time through what he then called “hot potato routing” and what we later called “packet-switching” and more recently “smart” protocols.31 A dynamic distributed network design, he thought, could survive the destruction of many nodes.
This survivalist logic is a good example of a larger truth: just because the networks he designed worked does not mean that networks work today because of those design values. As in the Soviet and Chilean cases, none of Baran’s stated design values (neither a survivable national security communication platform nor cybernetic neural-national networks) cash out in obvious historical consequences in the long run. Instead, the institutions at hand played a huge role in shaping—and in part delaying—the actual development of the ARPANET. Baran faced resistance from upper management that balked at the cross-agency collaborations such a network would require. For example, Baran’s supervisors in government and industry ignored his 1964 research report “On Distributed Communication” for several years. It appears likely that Baran’s supervisors deprioritized his work because it did not have weight of the top-secret imprimatur: if so, it is bitter comment on how interagency collaboration often bloomed in secrecy, not open, cultures, since Baran deliberately chose to publish his research publicly in the hopes his designs would reach the Soviets and ensure mutual competitiveness. (Perhaps the limits of contemporary open-source politics—such as the radical transparency of WikiLeaks and the private cipher that is Julian Assange—follow the limits of Cold War mutual intelligibility.) AT&T, the near analogue to a state network monopoly at the time, also rejected Baran’s proposal to develop digital networks, fearing digital networks would bring them into competition against their own services.32
And so Baran’s research languished until an Englishman named Donald Davies, with the support of the UK post office, independently developed network innovations similar to Baran.33 Only then, in 1966, under the pressure of an outside organization to which the national government felt compelled to respond (an ally nation backed by another hefty state media monopoly), did Baran’s superiors feel pressured to revisit and reclaim his team’s network research for ARPA under J. C. R. Licklider. The fact that we know networks now as “packet switching” (Davies’s term), not “block switching” (Baran’s term), reminds us that the realities of collaboration take shape independent of, and even in spite of, their design values. Accident and contingency coauthored the internet.
Network history takes shape apart from its most cherished design values. As the Soviet and Chilean insights add to the well-known American story, the conditions for building and operating networks rest on the tricky politics of convincing institutions to collaborate and make vital infrastructural investments. The design values that retroactively appear to justify the American case should not do so. Large-scale institutions—like states and global corporations—must do the hard work of coming to agree on how to build vital infrastructure projects, including networks—but what a network looks like on paper is not among them. For the most part, Cold War states and corporations have collaborated when pressured into imitating their enemies (who are often already imitating them). The historical fickleness of institutional collaboration is not encouraging news for any modern romance with etwork design.
Consider two values that network history cannot be attributed to. First, the belief in distributed network design—that networks should connect each node to its neighboring nodes in a democratic ideal free of bottlenecks—appears to be one of the most enduring and popular network myths. As of writing, the term “distributed network,” for instance, garners ten times more hits on Google than “decentralized networks,” despite the startlingly simple empirical observation that large-scale distributed networks do not, in fact, exist in the real world. No significant computer network today is distributed, and the internet has never been distributed: when the ARPANET had only four nodes, one node (in Utah) had only one severable link to the other three nodes (in California). It was never evenly connected to all its peer nodes.
Some nodes matter more than others. In practice, all existing major computer networks today are unevenly decentralized, meaning that all large-scale computer networks come with consequential (but often invisible) pinch points, veto sites, and obligatory passage points and clusters. In 2011, Egyptian President Hosni Mubarak, for example, needed to pressure only a handful of internet service providers to turn off the internet in Cairo during the Arab Spring.34 A single anchor dragged along an ocean floor pushed Somalia off the grid; the shovel of a seventy-five-year-old woman scrounging for spare copper caused an internet blackout across Armenia. The uneven material realities behind communication networks today make mincemeat of the democratic design ideals baked into increasingly global technology. Or rather, and perhaps more humbling still, networks show that, like actually existing democracy, our network realities are far from ideal and require continuous maintenance and attention.35
Second, internet packet-switching protocol, once heralded by its innovators as a solution to the problem of hierarchy, no longer seems the solution it once did. These protocols have been celebrated for smartening peer-to-peer, end-to-end communication across dumb networks for more than thirty years—that’s not necessarily wrong, although in the same period the protocols have grown too smart: the consequences of fragile bottlenecks in decentralized networks pale in comparison to the fact that now every node in a network can be leveraged and exploited through smart backdoors built into recent editions of the internet protocol. The internet protocol works by letting each node make microdecisions about how to reroute packets in real time; however, as media scholar Florian Sprenger has argued, the microdecisions protocols have let corporate and state spies introduce deep-packet inspection and off-the-record access points to each node, shepherding in mass network surveillance.36 Network technologies now permit the simultaneously location and addressing of mobile devices and their users worldwide. This means that surveillance networks are not the territory they map, nor, as these histories show, are networks the map itself.37 The fact issues a stirring wake-up call: the protocol once celebrated as solving the problem of hierarchical control has proven the very vehicle for ushering in the current era of private network surveillance. Internet protocol, which Baran once diminished as just “a few housekeeping functions such as standards setting,” now display, as in the Roman Saturnalia festivals where slaves would become masters, the hidden power of the housekeeper.38
The historical life cycle of global computer networks is perhaps in its adolescence (and, as I suggest elsewhere in this volume, midlife crisis). In its childhood, many tech visionaries called networked new media fundamentally different from conventional mass media of the twentieth century; yet future historians will likely frame the mass-mediated twentieth century as a prelude to, not a break from, the mass-networked twenty-first. Most users and technologists would do well to check rose-colored visions of our ever-brighter network future with a balance of self-critical reflection and risk-adverse maturation. This essay has taken a step in that direction by outlining three design-defiant early national computer networks projects—the Soviet OGAS, the Chilean Cybersyn, and the US ARPANET. Throughout I have argued that every history of a computer network is first a history of the organizations that tried to build it—and only secondarily a reminder of our collective failed romance with their design values. Table 3.1 summarizes the differences between what network designers thought they were doing (stated design values, left column) and what their institutional practices made happen (right column).
Table 3.1 Stated Design Values and Institutional Practices in Three Early Computer Networks
| Stated Design Values | Institutional Practices | |
|---|---|---|
|
Soviet OGAS |
Socialist economic reform |
Official consensus on purpose |
|
Nation as cybernetic body |
Competitive in-fighting |
|
|
Three-tiered hierarchy plan |
No centralized funding |
|
|
Explicit state surveillance |
Bureaucratic heterarchy |
|
|
Chilean Cybersyn |
Socialist economic reform |
Official consensus on purpose |
|
Nation as cybernetic body |
External conflict and revolt |
|
|
“Viable system” design |
Unviable political environment |
|
|
Class equality through male power |
||
|
US ARPANET |
Surviving a nuclear attack |
Ambiguity and conflicting purposes |
|
Nation as cybernetic brain |
Collaborative research |
|
|
Scientific data exchange |
State funding of nondefense research |
|
|
Global mass surveillance |
In all three cases, the stated design values in the left column approximate how these network projects are remembered, while the right column better describes how the networks actually developed (or not) over time. Several other observations arise. Midcentury cybernetic system analogies inspired all three networks, although never to a predictable effect. So while one may be tempted to, say, celebrate or critique the ARPANET for its design commitment to incorporating individual citizens as thinking nodes in a national brain, not working cogs in the (Soviet or Chilean) national body, these easy glosses fall short of the mark: network history followed neither. The beautiful design symmetries between technology and social values are more likely to mislead than to be realized. Meanwhile, the features of network history—from the technical affordances of network protocols that permit surveillance backdoors to the historical unrest of middle-class revolutionaries, to funding squabbles, to military-industry stonewalling—will continue to unfold with a forceful reality no design symmetry can imagine or manage. Modern global networks arrived on the backs of institutional ambiguities, nonhierarchical (or, borrowing from McCulloch, what I call “heterarchical”) conflicts of interest, and the tumult of socioeconomic unrest.39 These powerful and contingent historical forces, once accounted for, help network observers resist the Narcissus temptation to see in networks only the rippled reflections of our own favorite design values. No understanding of the organizations that build networks will cohere so long as we insist that every design is necessary, not contingent; the organizational mode we call a “network” is itself a (often too) convenient fiction built out of other organizational contingencies.
If users and designers must espouse values, let us lead with risk-averse insights into their vices, not virtues. Indeed, a healthy network public might prefer the Soviet situation in which its state explicitly sought to build a surveillance network (left column Soviet) to that of the American, which only discovered the surveillance baked into its networks after the fact (right column US). Today’s euphemism “cloud computing,” a public relations triumph taken straight from the playbook of immaterial media myths, lacks the punch and clarity with which the Soviet network designers described “remote-access networking”; at least the Soviets had the pluck to call cloud computing what it remains today—a network designed to monitor, track, and survey all nodes in its reach.40
The network vocabulary inherited from the Cold War also needs a serious update: the oppositions between capitalism and socialism, private corporation and public state do not adequately describe the complex institutional practices that supported computer networks in Cold War mixed economies or today’s global network economy. Understanding the Cold War as an ideological battle between social freedoms and fairness, it is clearer in the wake of Edward Snowden’s revelations of the NSA’s global surveillance and the Russian hacking of the 2016 US elections that the very technology thought to herald the end of that conflict and to usher in an age of information freedom—global computer networks—has actually left no clear victor among our multipolar world of networked citizens and migrants.41 Instead, the sure victors of the last half century of network history have been large organizations—telecommunication companies, surveillance states, big tech (Amazon, Apple, Facebook, Google), internet providers, cybersecurity organizations, data broker services (SAS and SalesForce), and other technologically powerful few that, fueled by data breaches and the fumes of modern privacy, sort, toll, and monitor our many network traffic bottlenecks.
Networks do not resemble their designs so much as they take after the organizational collaborations and vices that tried to build them: the first civilian computer networks ever imagined tried to reproduce and manage the institutional collaboration among Soviet economic bureaucracy—and when Soviet bureaucrats proved noninteroperable, so did their computer networks. Meanwhile, the hopes of the first Chilean national network project aspired to high cybernetic theory in design, even as they were dashed against the hard soil of class history and ideological unrest. Today, the internet still reproduces many of the institutional ambiguities that have long been baked into it: while the internet can still appear generative and open to the individual user, it has, since the Department of Defense first dreamed it up, scaled to serve unscrupulous institutions. Instead of circumventing institutional power hierarchies, computer networks have embedded themselves into, served, reconfigured, and then largely reconstituted the hierarchies of the Department of the Defense. (Even open-source cultures celebrate and elevate celebrity authorities.) The early success of the ARPANET rests on the collaborations between the dashes of the US military-industrial-academic complex in the 1960s—and now, again, the internet serves first the interests of nonpublic corporate and state surveillance. The “mother of all nets,” as the patriarchs of the ARPANET dubbed it, has since seeded hundreds of thousands of closed, proprietary computer networks. So too has the internet sped a media environment that rewards paywalls, firewalls, and the walled gardens we call apps, not because that design is embedded in the technology but because such tools are fashioned in the imprint, not the image, of global capital, network power, and human greed.
Given all this, we need new kinds of network organization collaborations—perhaps not unlike fiduciaries in medicine—whose self-interests are checked and balanced and whose fundamental aim is not fulfilling imagined design values but maintaining the health of the world it networks.42 Until then, the internet now encircling the world—which is neither distributed nor evenly decentralized—is first among many puckish and powerful muses that the industrialized West built on the fortunes of its own misbehavior. But rarely by design. Given the widening gaps in our understanding of the unruly global history of networks, it is a safe bet that, until checked, networks and their designs will double-cross our fondest wishes time and again. Until then, a network is not its map. A network is anything but a network.
1. Craig Timberg, “Net of Insecurity: A Flaw in Design,” Washington Post (May 30, 2015), https://www.washingtonpost.com/sf/business/2015/05/30/net-of-insecurity-part-1/.
2. Manuel Castells, Rise of the Network Society and End of the Millennium. in his trilogy The Information Age—Economy, Society, and Culture (Oxford: Blackwell, 1998); Bruno Latour, Reassembling the Social: An Introduction to Actor-Network-Theory (Oxford: Oxford University Press, 2005); Geert Lovink, Networks without a Cause: A Critique of Social Media (New York: Polity, 2012).
3. The Soviet case study builds on, among others, my most recent book, How Not to Network a Nation: The Uneasy History of the Soviet Internet (Cambridge, MA: MIT Press, 2016).
4. The Chilean case study builds on the authoritative work that is Eden Medina’s Cybernetic Revolutionaries: Technology and Politics in Allende’s Chile (Cambridge, MA: MIT Press, 2011).
5. Janet Abbate, Inventing the Internet (Cambridge, MA: MIT Press, 1999); Paul N. Edwards, The Closed World: Computers and the Politics of Discourse in Cold War America (Cambridge, MA: MIT Press, 1996); Finn Burton, Spam: A Shadow History of the Internet (Cambridge, MA: MIT Press, 2013); Thomas Streeter, The Net Effect: Romanticism, Capitalism, and the Internet (New York: NYU Press, 2011). See also Katie Hafner and Matthew Lyon, Where Wizards Stay Up Late: The Origins of the Internet (New York: Simon & Schuster, 1996) and Walter Isaacson, The Innovators: How a Group of Hackers, Geniuses, and Geeks Created the Digital Revolution (New York: Simon & Schuster, 2014).
6. Peters, How Not to Network a Nation, 1–13. See also Slava Gerovitch, “InterNyet: Why the Soviet Union Did Not Build a Nationwide Computer Network,” History and Technology 24, no. 4 (2008): 335–350.
7. Francis Spufford, Red Plenty (Minneapolis: Greywolf Press, 2010).
8. Aleksandr Ivanovich Stavchikov, academic secretary of CEMI, “Romantika pervyikh issledovanii I proktov I ikh protivorechnaya sub’dba” [Romanticism of early research and projects and their contradictory fate], unnamed, unpublished history of Central Economic Mathematical Institute (CEMI), Moscow, read in person and returned May 2008, chapter 2, 17.
9. Peters, How Not to Network a Nation, 108–114.
10. Quoted in Peters, How Not to Network a Nation, 119.
11. Peters, How Not to Network a Nation, 18, 27, 55, 95–96, 100, 118–120, 202.
12. Gertrude Schroeder, “The Soviet Economy on a Treadmill of Reforms,” Soviet Economy in a Time of Change, US Congress Joint Economic Committee (Washington, DC: USGPO, 1979).
13. Peters, How Not to Network a Nation, 63, see also the rest of chapter 2.
14. Peters, How Not to Network a Nation, 160–166.
15. Peters, How Not to Network a Nation, 10.
16. Medina, Cybernetic Revolutions.
17. Stafford Beer, Brain of the Firm (London: Penguin Press, 1972); Stafford Beer, Heart of the Enterprise (Chichester: Wiley, 1979); Stafford Beer, Diagnosing the System for Organizations (New York: Wiley, 1985). See also, Medina, Cybernetic Revolutions, 34–39, 121, 143.
18. Medina, Cybernetic Revolutions, 45, 93.
19. Eden Medina, “The Cybersyn Revolution: Five Lessons from a Socialist Computing Project in Salvador Allende’s Chile,” Jacobin (April 27, 2015), https://www.jacobinmag.com/2015/03/allende-chile-beer-medina-cybersyn/.
20. Medina, “The Cybersyn Revolution.”
21. Medina, “The Cybersyn Revolution.”
22. Medina, “The Cybersyn Revolution.”
23. Peters, How Not to Network a Nation, 200.
24. Streeter, The Net Effect.
25. See Abbate, Inventing the Internet.
26. Robert A. Divine, The Sputnik Challenge: Eisenhower’s Response to the Soviet Satellite (New York: Oxford University Press, 1993); Audra J. Wolfe, Competing with the Soviets: Science, Technology, and the State in Cold War America (Baltimore: John Hopkins University Press, 2012); Annie Jacobsen, The Pentagon’s Brain: An Uncensored History of DARPA, America’s Top Secret Military Research Agency (New York: Little, Brown, 2015); Peters, How Not to Network a Nation, 92; Stuart W. Leslie, The Cold War and American Science: The Military-Industrial-Academic Complex (New York: Columbia University Press, 1993), 203–231.
27. I am not aware of much sober scholarship on this particular transition from East Coast government to Silicon Valley private business, although much of the dated rhetoric that pits state against corporations can be found in popular accounts, such as Michael Hiltzik, Dealers of Lightning: Xerox PARC and the Dawn of the Computer Age (New York: Harper Business, 1999), and L. Gordon Crovitz, “Who Really Invented the Internet?” Wall Street Journal (July 22, 2012), https://www.wsj.com/articles/SB10000872396390444464304577539063008406518.
28. Judy O’Neill, “Interview with Paul Baran,” Charles Babbage Institute, OH 182 (February 5, 1999), Menlo Park, CA, accessed September 15, 2017, http://www.gtnoise.net/classes/cs7001/fall_2008/readings/baran-int.pdf.
29. Tara Abraham, Rebel Genius: Warren S. McCulloch’s Transdisciplinary Life in Science (Cambridge, MA: MIT Press, 2016).
30. Stewart Brand, “Founding Father,” Wired 9, no. 3 (1991), http://archive.wired.com/wired/archive/9.03/baran_pr.html.
31. Paul Baran, “On Distributed Communications: II. Digital Simulation of Hot-Potato Routingin a Broadband Distributed Communications Network,” RAND Publications, 1964, accessed September 15, 2017, https://www.rand.org/pubs/research_memoranda/RM3103.html.
32. Peters, How Not to Network a Nation, 96–97.
33. Hafner and Lyon, Where Wizards Stay Up Late, 64.
34. Larry Greenemeir, “How Was Egypt’s Internet Access Shut Off?,” Scientific American 28 (January 2011), https://www.scientificamerican.com/article/egypt-internet-mubarak/.
35. Tom Parfitt, “Georgian Woman Cuts Off Web Access to Whole of Armenia,” Guardian (April 6, 2011), https://www.theguardian.com/world/2011/apr/06/georgian-woman-cuts-web-access.
36. Florian Sprenger, The Politics of Micro-Decisions: Edward Snowden, Net Neutrality, and the Architectures of the Internet (Lueneburg, Germany: Meson Press, 2015). See also Alex Galloway, Protocol: How Control Exists after Decentralization (Cambridge, MA: MIT Press, 2004); Laura Denardis, Protocol Politics: The Globalization of Internet Governance (Cambridge, MA: MIT Press, 2009).
37. Florian Sprenger, “The Network Is Not the Territory: On Capturing Mobile Media,” New Media & Society 21, no. 1 (2019): 77–96.
38. Paul Baran, “Is the UHF Frequency Shortage a Self Made Problem?” Accessed September 15, 2017, http://digitallibrary.usc.edu/cdm/ref/collection/p15799coll117/id/2494; also quoted in Galloway, Protocol, 120.
39. David Stark, The Sense of Dissonance: Accounts of Worth in Economic Life (Princeton, NJ: Princeton University Press, 2009). See also Peters, How Not to Network a Nation, 22–24, 145, 173, 193.
40. Peters, How Not to Network a Nation, 109, 112. See also John Durham Peters, The Marvelous Clouds: Toward a Philosophy of Elemental Media (Chicago: Chicago University Press, 2015), and Vincent Moscow, To the Cloud: Big Data in a Turbulent World (New York: Paradigm Publishers, 2014).
41. See Mario Biagioli and Vincent Lepinay, eds., From Russia with Code: Programming Migrations in Post-Soviet Times (Durham, NC: Duke University Press, 2019).
42. Jack M. Balkin, “Information Fiduciaries and the First Amendment,” UC Davis Law Review 49, no. 4 (2016); see also Jack M. Balkin and Jonathan Zittrain, “A Grand Bargain to Make Tech Companies More Trustworthy,” Atlantic (October 3, 2016); see also Lina Khan and David Pozen, “A Skeptical View of Information Fiduciaries,” Harvard Law Review 133 (May 25, 2019), https//ssrn.com/abstract=3341661.