Dealers of Lightning Read online

  Clark thought computer science would be better served by jumping directly to single-user machines, even if that meant temporarily mak­ing do with underpowered computers. "He would talk about how it was not going to be too many years before we would have a computer you could hold in your hand," recalled Severo Ornstein, a PARC engi­neer who was one of Clark's longtime associates. "At that time comput­ers were filling buildings larger than this one—a single computer. But he said, 'Yeah, you'll just paint 'em on your desk, just like that.' So a lot of us felt that time-sharing was an enormous waste."

  Starting in 1962 Clark underscored his conviction by designing and building the legendary "LINC." (Its name stood for "Laboratory Instru­ment Computer" but echoed the name of its birthplace, MIT's Lincoln Lab.) The LINC was unique for its time in that it could be operated by a single user from a desk-sized console, aldiough its processor and memory were housed in a wardrobe-sized unit typically concealed in a nearby closet. Designed specifically to serve biomedical research rather than as a general-purpose machine—which helped keep it compact—the LINC "was the first machine that you could take apart and put in the back of your car, carry somewhere else, put back together again, and it would run," Ornstein recalled. "That idea had never previously seemed con­ceivable."

  But the machine was almost too fascinating. It attracted the interest of the National Institutes of Health, which in 1964 offered MIT the unprecedented sum of $37 million to establish around the LINC an inter-university program of computer-aided biomedical research. The scent of money attracted MIT's academic mandarins to a project they had previously relegated to the fringes of the research departments, which Clark preferred anyway. He did not relish seeing his own pro­gram coming under the academic establishments thumb. Obstreper­ous to the last, he flatly refused to cooperate, forcing the exasperated university to abruptly withdraw its support for the entire venture.

  For the next few weeks Clark frantically canvassed the country to find a new home for the machine and the dozen junior researchers whose livelihoods and careers depended on it. Washington University, which was trying to build both a medical school and a digital computing pro­gram, saw opportunity in MIT's pique, not to mention the chance to turn the tables on a big East Coast institution by raiding it for a change. A few days before his scheduled eviction from Lincoln Lab, Clark looked up from his desk. A stranger in spectacles was standing in the office, stam­mering out a transparent stoiy about "just happening" to be passing through Cambridge en route to an engagement in Woods Hole. It was George Pake, come to check Clark out on the recommendation of a mutual friend. Before returning to St. Louis a few days later, he agreed to give Clark's project a permanent home at Washington University, where it was to obtain further funding through one Bob Taylor, at ARPA.

  Pake and Taylor each came away from this initial interaction favorably disposed toward the other. Pake was impressed by Taylor’s excellent contacts within the computing fraternity and his apparent authority to disburse millions of dollars with a minimum of fuss. (Formally speak­ing, Taylor was still Ivan Sutherland's deputy at the time.) Taylor saw Pake as a pragmatic administrator capable of cutting through red tape to assist a program and a researcher he valued highly. They obviously could have had no inkling of how, within a few short years, their lives would intertwine as colleagues and adversaries.

  Before the two would have a chance to meet again Taylor's capacious net would come to embrace areas of computer research that barely existed when ARPA delivered its lifesaving shot to Wes Clark's project. At ARPA he funded the country's first full-fledged graduate degree programs in computer science at Stanford, Carnegie-Mellon, and MIT. Some fields of study virtually owed their existence to his largesse. Among them was computer graphics, which came to life at the Univer­sity of Utah when Dave Evans, a devout Mormon who had led the Genie team building the time-sharing SDS 940 at UC Berkeley, called Taylor to say his alma mater had invited him to return to Salt Lake to start a computer program. How about an ARPA project, he asked, to get it going?

  Computer graphics was then attracting almost no one's attention, for the simple reason that most computers lacked visual displays of any kind. If Evans was willing to start such a program in the backwater of Utah, where it could develop in pristine isolation from the traditionalist think­ing elsewhere, Taylor was all for it. The venture turned out better than anyone could have expected. The program Taylor funded partially as a personal experiment and partially as a favor to an old friend evolved into a world leader in computer graphics research.

  His most enduring legacy, however, was not a university program but a leap of intuition that tied together everything else he had done. This was the ARPANET, the precursor of today's Internet.

  Taylor's original model of a nationwide computer network grew out of his observation that time-sharing was starting to promote the formation of a sort of nationwide computing brotherhood (at this time very few members were women). Whether they were at MIT, Stanford, or UCLA, researchers were all looking for answers to the same general questions. "These people began to know one another, share a lot of information, and ask of one another, 'How do I use dais? Where do I find that?'" Taylor recalled. "It was really phenomenal to see this computer become a medium that stimulated the formation of a human community."

  There was still a long way to go before reaching that ideal, however. The community was less like a nation than a swarm of tribal hamlets, often mutually unintelligible or even mutually hostile. Design differ­ences among their machines kept many groups digitally isolated from the others. The risk was that each institution would develop its own unique and insular culture, like related species of birds evolving inde­pendently on islands in a vast uncharted sea. Pondering how to bind them into a larger whole, Taylor sought a way for all groups to interact via their computers, each island community enjoying constant access to the others' machines as though they all lived on one contiguous vir­tual continent.

  This concept would develop into the ARPANET. The idea owed something to Licklider, who had earlier proposed what he dryly called an "intergalactic network" of mainframes. During his time at ARPA the notion remained theoretical, however; it was hard enough to get small-scale time-sharing systems to run individually much less in con­cert with one another. But Taylor judged that the technology had now progressed far enough to make the concept practical. He did not deceive himself: Building such a system meant overcoming prodigious obstacles. On the other hand, ARPA's generous umbrella sheltered hundreds of scientists and engineers whose prodigious talents, he rea­soned, were fully up to the challenge.

  One day in February 1966 Taylor knocked at the office of ARPA's director, the Austrian-born physicist Charles Herzfeld, armed with lit­tle more than this vague notion of a digital web connecting bands of time-sharers around the country. At any other agency he would have been expected to produce reams of documentation rationalizing the program and projecting its costs out to the next millennium; not ARPA. "I had no formal proposals for the ARPANET," he recounted later. "I just decided that we were going to build a network that would connect these interactive communities into a larger community in such a way that a user of one community could connect to a distant commu­nity as though that user were on his own local system."

  After listening politely for a short time. Herzfeld interrupted Taylors rambling presentation. He had followed his young associate s theoreti­cal research closely enough to know already the gist of his ideas. All he had was a question.

  "How much money do you need to get it off the ground?"

  "I'd say about a million dollars or so, just to start getting organized."

  "You've got it," Herzfeld said.

  "That," Taylor remembered years later of the meeting at which the Internet was born, "was literally a twenty-minute conversation."

  Actually getting the program underway required some further maneu­vering, Taylor-style. His candidate for program manager, a twenty-nine- year-old MIT researcher named Lawrenc
e G. Roberts, refused to leave his secure and intellectually rewarding post at Lincoln Lab despite Tay­lor's relentless wheedling. After seven or eight months, Taylor was des­perate to resolve the standoff.

  "Do we still support fifty-one percent of Lincoln Lab?" he asked Herzfeld, who confirmed the figure. Taylor asked Herzfeld to put in a call to Lincoln's director. "Tell him that it's in Lincoln Lab's and ARPA's best interests to tell Larry Roberts to come down and do this." Within two weeks, Roberts accepted a job that would eventually secure him a permanent place in the computing Pantheon, as the Internet’s found­ing engineer. As Taylor later crowed: "I blackmailed Larry Roberts into fame!"

  But by 1969 Bob Taylor was feeling burned out. He had spent more than four years at ARPA's Information Processing Technologies Office, nearly three of those as director. His annual research budget of $30 million had become the single most important force in U.S. computer research. But the research agency was changing around him. The inescapable catalyst was Vietnam.

  In 1967 the war had reached into the comfortable civilian enclosure of ARPA and touched Taylor personally. The Johnson White House had appealed for help with a logistical nightmare that had nothing to do with materiel or troop deployment. The issue was information. The Vietnam military command, it seemed, had got itself bogged down in a statistical quagmire. "There were discrepancies in the reporting com­ing back from Vietnam to the White House about enemy killed, sup­plies captured, bullets on hand, logistics reports of various kinds," Tay­lor recalled. "The Army had one reporting system; the Navy had another; the Marine Corps had another."

  Unsurprisingly, this system produced ludicrous results. Estimates of enemy casualties exceeded the known population of North Vietnam, while the reported quantities of captured sugar reached levels equiva­lent to three-quarters of the world supply. "It was ridiculous. Out of frustration the White House turned to the Secretary of Defense to clean this mess up. The Secretary of Defense turned to ARPA, because ARPA was a quick-response land of agency. The director of ARPA asked me to go out to Vietnam and see whether or not any land of computer technology could bring at least some semblance of agree­ment, if not sanity, to this whole process."

  Joined by his assistant, Barry Wessler, and three Pentagon-based rep­resentatives from the Army, Navy, and Air Force, Taylor made several trips to the war zone. The situation was even worse than he expected. The military was literally drowning in information. Data flowed into depots and never flowed out. Pilots returning from missions would get debriefed their reports entered on punch cards; then their co-pilots would get debriefed and their reports recorded. But no one did anything with the information, which piled up without anyone bothering to figure out how or even why these reports should be collated and organized.

  Taylor assigned technical teams to the trouble spots to straighten out the chaos, although not without meeting resistance. Occasionally some base commander would refuse to grant ARPA's civilian analysts access to his precious cache of useless data, at which point Taylor, who trav­eled on government business as a one-star general, would be forced to step in and pull rank.

  Taylor and his group solved the military's problem, after a fashion. They installed a master computer at the U.S. military command head­quarters at Ton Son Nhut Air Base and made it the lone repository of all data. "After that the White House got a single report rather than several," Taylor remarked. "Whether the data were any more correct or not I don't know, but at least it was more consistent."

  But the experience left him feeling increasingly uneasy about his role at the Pentagon. "My first trip out to Vietnam I was thinking, 'Well, we're doing a good thing for these oppressed people. We're out here to clean this mess up.' But by the second or third trip I realized this is a civil war and I didn't want to have much to do with it. Nor did I think my country should have anything to do with it."0

  Adding to his frustration was the war's increasing toll on ARPA. For most of the decade the agency's civilian character had insulated it from the deepening rifts within the military establishment. But as the war encroached more and more, the agency had to fight for resources. By the close of the 1960s the Pentagon had slashed ARPA's budget to half of what it had been at mid-decade.

  The agency faced mounting political troubles, too. The notion that any arm of the Pentagon could engage in wholly innocent and purely civilian research incited mistrust across the country. As a defensive measure, ARPA started to shed its civilian entanglements and consciously remake itself into what the nation thought it was anyway—an arm of the war machine. When the Caltech engineer Eberhardt Rechtin succeeded Herzfeld as director in 1967, he assured his congressional overseers he would nudge ARPA toward "mission-oriented" objectives—programs aimed at satisfying chiefly military goals. The 1969 Mansfield Amend-

  "Perhaps he was also put off by the effect American morals and money were having on the bucolic country. Wessler recalled an incident one evening when he and Taylor were being relentlessly importuned by two Vietnamese prostitutes at the bar of their Saigon hotel. One pressed herself with particular vigor on Taylor, who kept turning her away with the excuse that as a mere government employee he could never meet her price. As the two men were leaving, the second prostitute stopped Wessler. "My friend would like to sleep with your friend," she said. "Would you please arrange it?" Wessler solemnly shook his head. "I do a lot of things for Bob Taylor," he replied. "But I don't do that." ment to the military appropriations bill would formalize the trend, direct­ing the Pentagon henceforth to fund only projects of obvious military rel­evance. As if to underscore the point, the amendment changed ARPAs name to DARPA, the Defense Advanced Research Projects Agency.

  Taylor beheld the emasculation of government research with frank alarm. "Most of the time I was there ARPA was going for projects with an order-of-magnitude impact on the state of the science," he reflected. "We had made a decision that we would not go for incre­mental things. But as soon as you get mission orientation you're focus­ing on very narrow objectives."

  He felt it was the right time to leave. He had held his job longer than both his predecessors combined. The ARPANET was securely launched under Larry Roberts's unwavering eye. In September the network's first four nodes—at UCLA, Stanford Research Institute, the University of California at Santa Barbara, and the University of Utah—went opera­tional. Taylor accepted an invitation from Dave Evans to help Utah undertake a research coordination effort of conveniently vague scope. In late 1969 he left Washington for good and headed for Salt Lake City. He was still there a year later when George Pake tracked him down.

  "I heard through the grapevine that he wasn't altogether happy at Utah," Pake recalled. This was certainly the prevailing opinion among Taylor's friends, who found it hard to imagine him careening through the stolid precincts of Salt Lake in his blue Corvette. On campus his recommendations to cancel some programs and merge others together caused, he freely admitted, "some dissatisfaction." Clearly Dave Evans had done him a favor by facilitating his departure from Washington. But beyond that, Bob Taylor was marking time.

  Pake's purpose in inviting Taylor to Palo Alto was to pick his brains rather than offer him a job (although he did not rule out the latter pos­sibility). He had been unable to solve his most pressing administrative problem: identifying the best researchers in the computing field. It was one thing to compile a list of the country's best computer science programs—the same few names kept coming up, including Berkeley, MIT, Carnegie-Mellon, and Stanford—but quite another to appraise the individual talents within, or to know which projects pointed toward progress and which were intellectual cul-de-sacs. Pake recognized that Taylor's job for five productive years had involved making exactly those sorts of judgments.

  A few days later Taylor was ushered into Pake's office on Porter Drive. There were two men in the room other than his host—Frank Squires, the personnel chief, and Bill Gunning, a pleasant and unas­suming engineer with twenty years' experience in analog and digital electronics
who had been appointed manager of PARC's Systems Sci­ence Lab.

  "They sat me down and Pake said, "We bought a computer company,'" Taylor recalled. "I said, 'Yeah, that's too bad. You bought the wrong one.' I told them that SDS wasn't interested in interactive computing, and that's what I'd be doing." Without humility he proceeded to summarize how he believed SDS and Max Palevsky had gone astray, the memory of his bitter encounter with the computer magnate (then still a Xerox director) apparently still fresh. As his hosts listened patiently, he outlined his vision of a future in which interactive computers harnessed to nationwide networks enhanced the communication of human to human.

  Pake, for one, took his caustic critique of SDS in stride. He had already encountered the people in El Segundo and largely agreed with Taylor's assessment. The lecture on distributed personal computing was a different matter. No one in the room valued Bob Taylor as an important theoretician and his digression elicited only their mental shrugs. "We were interested in him not because of any vision he had of distributed computing," recalled Squires, "but because of the people he knew—and that meant every significant computer scientist in the United States."