by Yasha Levine
Stephen Wolff’s job at Aberdeen in the 1970s involved working on the ARPANET and linking it with the US Army’s network of supercomputers.39 In 1986, the National Science Foundation’s Networking Office hired him to do the same thing, but with a major twist: he was to build a government-funded network that extended the ARPANET design into the civilian world, and then spin this network off to the private sector.40 In the end, Wolff oversaw the creation and privatization of the Internet.
When I spoke to Wolff, I asked, “Is it right to call you the man who privatized the Internet?”
“Yes, that is a fair assessment,” he replied.41
Even before Stephen Wolff arrived at the National Science Foundation, it was clear ARPANET’s days were numbered. In 1975, the Pentagon had officially relieved ARPA of its responsibilities for running the network and placed it under the direct control of the Defense Communication Agency. The army, navy, air force, and National Security Agency had all started building out their own networks based on ARPANET technology. They maintained links to the original ARPANET infrastructure, but the physical network, with its limited 56K modem speeds, was beginning to show its age. The experiment had been a success, but as the 1980s approached, it looked as if the original ARPANET was going to be dumped in the trash.
The network had become obsolete, but the technology and framework on which it ran were only getting started. Many of the original ARPANET architects and designers cashed in on their ARPA experience in trade for lucrative private sector jobs in the rapidly growing computer networking industry; others remained at the Pentagon, pushing and evangelizing for a wider adoption of the ARPANET network design. Many were eager to see the original ARPANET grow beyond military circles and into a commercial network that everyone could use.42 The National Science Foundation (NSF), a federal agency created by Congress in 1950 with a mission to “promote the progress of science” and “secure the national defense,” was the vehicle that would ultimately get the job done.
In the early 1980s, the NSF ran a small network that connected a handful of university computer science departments to the ARPANET. By 1985, administrators wanted to expand the project into a bigger, faster network that would connect a larger pool of universities, extending the ARPANET out of purely military and computer science circles and making it available to all academic and educational users.43 On the basis of his decade of experience connecting US Army supercomputers to the ARPANET at Aberdeen, Wolff was brought on to build and manage this new educational network project—called the NSFNET.
The first version of the NSFNET came online in 1986. It was a modest effort, connecting supercomputer centers at five universities funded by the NSF, wiring them together so they could share data, and plugging them into a wider set of universities connected to the old military ARPANET. The network was limited in scope, but demand for it was so high that it crashed the system. Its puny leased lines had the combined throughput of a slow modem and could not handle the swell of users. Clearly, the NSFNET needed a major upgrade and more bandwidth. The question was: What would this new network look like?
The answer came quickly.
“Starting with the inauguration of the NSFNET program in 1985, we had the hope that it would grow to include every college and university in the country,” recalled Wolff in an interview.44 “But the notion of trying to administer a three-thousand-node network from Washington—well, there wasn’t that much hubris inside the Beltway.”
Hubris, indeed. This was the height of the Reagan era, a time of privatization and deregulation, when public ownership of vital infrastructure was considered a barbaric relic that had no place in the modern world; if anything, it needed to be lanced like a boil. Everything was being deregulated and privatized—from the banking sector to telecommunications and broadcast industries. Wolff and his team at the NSF, like the obedient public servants that they were, toed the line.
In early 1987, he and his team finally hashed out a design for an improved and upgraded NFSNET. This new network, a government project created with public money, would connect universities and be designed to eventually function as a privatized telecommunications system. That was the implicit understanding everyone at NSF agreed on. They viewed the public nature of the NSFNET as a transitory state: a small government pollywog that would transition into a commercial bullfrog. According to specs, the new NSFNET would be built as a two-tier network. The top layer would be a national network, a high-speed “backbone” that spanned the entire country. The second layer would be made up of smaller “regional networks” that would connect universities to the backbone. Instead of building and managing the network itself, the NSF would outsource the network to a handful of private companies. The plan was to fund and nurture these network providers until they could become self-sufficient, at which point they would be cut loose and allowed to privatize the network infrastructure they built for the NSFNET.
Later in 1987, the NSF awarded contracts for its upgraded NSFNET design. The most important part of the system, the backbone, was run by a new nonprofit corporation, a consortium including IBM, MCI, and the state of Michigan.45 The second-tier regional networks were farmed out to a dozen other newly created private consortiums. With names like BARRNET, MIDNET, NYSERNET, WESTNET, and CERFNET, they were run by a mix of universities, research institutions, and military contractors.46
In July 1988, the NSFNET backbone went online, connecting thirteen regional networks and over 170 different campuses across the country.47 The physical network ran on MCI’s T-1 lines capable of transmitting 1.54 megabits per second and was routed through data switches built by IBM. The network stretched from San Diego to Princeton—snaking through regional network exchange points in Salt Lake City, Houston, Boulder, Lincoln, Champaign, Ann Arbor, Atlanta, Pittsburgh, and Ithaca and throwing out an international transatlantic line to the European Organization for Nuclear Research in Geneva.48 The network was a huge success in the academic community.49
Even as demand surged, NSF managers began the privatization process. “We told them: ‘You guys will eventually have to go out and find other customers. We don’t have enough money to support the regionals forever.’ So they did,” Wolff explained. “We tried… to ensure that the regionals kept their books straight and to make sure that the taxpayers weren’t directly subsidizing commercial activities. But out of necessity, we forced the regionals to become general-purpose network providers.”50
Telling NSFNET providers to diversify their client base by seeking commercial clients—it seems like a minor decision. Yet, it is a crucial detail that had a huge impact, allowing the agency a few years later to quietly and quickly privatize the Internet while making it seem like the transition was inevitable and even natural. People on the inside understood the gravity of what Wolff and the NSF were doing. They saw it as a kind of clever trick, a sleight of hand.
Vinton Cerf, who in 1982 had left his job at ARPA to head up MCI’s networking division, described Wolff’s private-public network provider scheme as “brilliant.” He said, “The creation of those regional nets and the requirement that they become self-funding was the key to the evolution of the current Internet.”51
Cerf is right. The Internet is perhaps one of the most valuable public inventions of the twentieth century, and decisions made by a few key unelected officials in the federal bureaucracy set the Internet on the certain path to privatization. There was no real public debate, no discussion, no dissension, and no oversight. It was just given away, before anyone outside this bureaucratic bubble realized what was at stake.
The privatization of the Internet—its transformation from a military network to the privatized telecommunications system we use today—is a convoluted story. Wade in deep enough and you find yourself in a swamp of three-letter federal agencies, network protocol acronyms, government initiatives, and congressional hearings filled with technical jargon and mind-numbing details. But on a fundamental level, it was all very simple: after two decades of lavish funding and re
search and development inside the Pentagon system, the Internet was transformed into a consumer profit center. Businesses wanted a cut, and a small crew of government managers were all too happy to oblige. To do that, with public funds the federal government created a dozen network providers out of thin air and then spun them off to the private sector, building companies that in the space of a decade would become integral parts of the media and telecommunications conglomerates we all know and use today—Verizon, Time-Warner, AT&T, Comcast.
But how did it happen exactly? Unravelling the tale requires looking at the first privatized NSFNET provider: a consortium led by IBM and MCI.52
The National Science Foundation functioned on an educational mandate and could support only initiatives that had an education dimension. Legally, NSFNET contractors were not allowed to route their commercial traffic through the government-funded network. These terms were baked into the federal agency’s “Acceptable Use Policy” contract, and they were quite clear. How could the network be privatized if it couldn’t route commercial traffic? Later, the NSF managers claimed that NSFNET providers didn’t violate these terms and that they routed commercial traffic through separate, privately built network infrastructure. But a backroom deal the NSF made with its backbone operator shows that the truth is a bit murkier.
In 1990, the MCI-IBM consortium, with approval from NSF, split into two corporate entities: a nonprofit called Advanced Network Services and a for-profit confusingly named ANS CO+RE Systems. Advanced Network Services—ANS—continued to contract with the NSF to maintain and run the physical NSFNET backbone. Meanwhile, its for-profit division, ANS CO+RE, sold commercial network services to business clients on a new network it called the ANSNET.53 Of course, this new ANSNET ran on the exact same physical network infrastructure that powered the NSFNET. Legally, though, the two—NSFNET and ANSNET—were treated as completely separate entities by the National Science Foundation, which meant that despite the Acceptable Use Policy that forbade commercial traffic on the NSFNET, the IBM-MCI consortium had a green light to do just that for profit.54 It was a clever maneuver. On a basic level, it allowed the MCI-IBM consortium to double book the same asset, pocketing government money to run the NSFNET and then selling this same network to commercial clients. More fundamentally, it allowed a corporate entity with a direct stake in the business of computer networking to privatize a government asset without doing so explicitly. That’s exactly how executives at MCI-IBM’s newly formed ANS division saw it: “[We] have privatized the NSFNET,” the president of ANS bragged at a networking industry workshop at Harvard in 1990.55
This public-private flip was not announced to the public, and it was also hidden from other NSFNET providers. When they finally discovered the existence of this sly deal a year later, they raised alarm and accused the agency of privatizing the network to a favored corporate insider. Some called for a congressional investigation into what they saw as mismanagement and possibly fraud. “It’s like taking a Federal park and giving it to K Mart. It’s not right,” a manager of a large NSFNET provider told the New York Times.56
They had a right to be upset. This backbone privatization deal gave a powerful company a privileged position that allowed it to quickly dominate the budding commercial networking market, frequently at the expense of other regional NSFNET providers.57 The key to this advantage was the NSFNET backbone itself. Built and sustained with government funds, the network spanned the width of the United States and had connections to more than thirty other countries. Regional networks, on the other hand, were smaller, usually restricted to geographic areas like Greater New York, the Midwest, or northern California. Those that expanded into the national commercial market could not route commercial traffic through the NSFNET backbone but had to build their own private networks without government funding. In short, the NSF directly subsidized the MCI-IBM consortium’s national business expansion. The company used its privileged position to attract commercial clients, telling them that its service was better and faster because it had direct access to the national high-speed backbone.58
Stephen Wolff understood that backing a telecommunications company like MCI could lead to a situation where a handful of powerful corporations controlled the newly created Internet, but he brushed those dangers aside. As Wolff explained in an interview at the time, his main objective was to bring a viable commercial Internet into existence. Regulating fairness and competitive practices was someone else’s job.59 On a very basic level, he was right. His objective was just to build a network, not regulate it. The problem was that by building a privatized network, he was also building an industry and, by extension, laying down the basic rules that governed and regulated this industry. These were intertwined.60
Wolff’s laissez-faire management style triggered an outcry among the smaller regional NSFNET providers. There were accusations of conflict of interest, insider dealings, favoritism. William Schrader, president of a New York area provider called PSINET, charged the NSF pointblank with granting a monopoly over government assets to a single privileged corporation. “The Government has privatized the ownership of a federal resource,” he said at a 1992 congressional hearing held to investigate possible government mismanagement of the NSFNET. “The privatization unnecessarily provided the contractor [IBM-MCI] with an exclusive monopoly position to use Federal resources paid by taxpayer funds.”61
Schrader’s PSINET banded together with other regional NSFNET providers to push the government to end MCI-IBM’s privilege and to finally open the network to unrestricted commercial traffic. “A level playing field can only be built by changing current NSF policies which favor one competitor,” Schrader told Congress.62
Schrader wasn’t contesting the privatization itself. Why would he? His own company, PSINET, had also been spun off from a regional NSFNET provider seeded with federal money as a for-profit entity.63 Like IBM-MCI’s ANS, PSINET represented a de facto privatization of a government-subsidized asset by a few privileged insiders who happened to be at the right place at the right time. Schrader didn’t challenge that. What he opposed was the NSF giving a different—and perhaps more powerful—group of privileged insiders more privilege than his company had received. This was a spat between competing government-subsidized networking companies in an industry created by the government. It was not a fight about privatization. It was a scrabble over how to divvy up the future profits in an emerging market worth billions.
In the mid-1980s, while Stephen Wolff was planning the NSFNET upgrade, the United States was in the grips of two closely related computer technology booms: the explosion of cheap personal computers and easy access to computer networking. First, IBM released a powerful personal computer and licensed the design so that any computer manufacturer could make compatible IBM computer components. A few years later, in 1984, Apple released the Macintosh, complete with a graphical user interface and mouse. Microsoft’s text-based DOS operating system for IBM computers was followed by a crude version of Windows. Computers were suddenly easy to use and affordable. It wasn’t just giant corporations, big universities, and government and military agencies anymore—smaller businesses and geeky middle-class early adopters could all get their own systems. It quickly became apparent that the true power of the personal computer was not personal at all, but social. Computers allowed people to tap into remote servers and connect with other computers, communicating and sharing information with people hundreds and thousands of miles away. Hundreds of thousands of people brought their computers home, plugged in their modems, and connected to a weird and early form of the Internet.
A few select companies had been providing ARPANET-like access to large corporations since the 1970s. But, in the late 1980s, all sorts of dial-up and networking services popped up across the country. There were big firms like CompuServe, Prodigy, and America Online as well as hundreds of smaller outfits. Some, no more than dial-up messaging boards, were run as hobbies on servers set up in basements and garages. Others were small businesses that
served up a series of features: forums, chat rooms, email, rudimentary computer games, and news.64 All of them were text-based and simple, a shadow of the real Internet that would emerge later, but they were extremely popular. Even Stewart Brand got onboard. He cofounded an early message board called The Well, which provided a forum and online meeting place for his vast network of hippie business associates, artists, writers, and journalists. The Well became popular very quickly, turning into a social hub for the up-and-coming “digirati”—Bay Area opinion makers, entrepreneurs, authors, hackers, and journalists who came to the fore in the 1990s to shape digital culture.
This was not the globally connected Internet we know today. Services like The Well and America Online were not connected to one another and allowed communication only between members of the same service. Effectively, they were siloed, at least for a time. Everyone in the industry understood that this was going to be a huge and extremely profitable industry, and that some kind of national network would connect it all. “It was no secret that whatever the network was then, it was going to become a big commercial success at some point. Nobody ever doubted that,” Wolff told me in an interview.65
Indeed, NSFNET contractors began fighting for control of this untapped and growing market as soon as Stephen Wolff gave them the green light to privatize their operations—that’s what the fight between providers like PSINET and ANS was all about. They were licking their chops, happy that the government bankrolled the network and even happier that it was about to get out of the business. There was a lot of money to be made. Indeed, by the end of the 1990s, Schrader’s humble PSINET had customers in twenty-eight countries and was worth $3 billion on the NASDAQ.66