BOB TAYLOR BOB TAYLOR
It started with a crash. On October 29, 1969, a $700,000 Sigma 7 computer at UCLA sent a command to a slightly leaner SDS 940 machine at the Stanford Research Institute.
The UCLA machine looked formidable. A half-dozen or more refrigerator-sized components lined the perimeter of a special room dedicated to their use, the entire operation controlled by an expert sitting at a typewriter-looking console in the center of the room.
But when the UCLA machine sent its command—LOGIN—up the California coast, the Stanford computer crashed before the word, typed letter by letter, even got past G.
After a bit of reprogramming, the message was sent again and received, and the first computer network, called the Arpanet, was online.1
This transmission, often hailed as the “birthday of the Internet,” has been celebrated in conferences, books, speeches, and news reports. Plaques have been erected in its honor. One man who worked on the network at UCLA has since recast the failed initial login in biblical terms: “And so the very first message ever sent over the Internet was ‘Lo!’ as in ‘Lo and behold!’ Quite a prophetic message indeed.”2
Of course, no such message was intended. That LOGIN was the computing equivalent of Alexander Graham Bell’s “Mr. Watson, come here”: a practical effort to determine if a message had been received. Whatever importance the LOGIN transmission has achieved by now, back in 1969, the sent message was not a momentous achievement. At UCLA and Stanford, there was a bit of applause and a lot of relief but no announcements from press offices, no reporters waiting to hear if the connection would work. A simple notation in a UCLA log book, “2230, Talked to SRI host to host,” served as recognition. At the universities—and at the Pentagon, where the Department of Defense had funded this new network; and in Cambridge, Massachusetts, where a small company called Bolt Beranek and Newman was building the equipment and writing much of the software—most people involved with the network scarcely paused to note the transmission.
And Bob Taylor, the man who had jump-started the network, paid no attention at all.
Three years earlier, in 1966, three thousand miles from Silicon Valley in Washington, DC, Taylor had been walking back from lunch with his new employee, twenty-three-year-old Barry Wessler. To be more exact: Taylor was walking, and Wessler was sneaking in the occasional half jog to keep up. Taylor—thirty-four, slight, a cloud of pipe smoke obscuring a face that reminded more than one person of John F. Kennedy—did nothing at a sedate pace. Every morning Taylor squealed his rare BMW 503 into the giant Pentagon lot after driving as fast as possible from his home in suburban Maryland. He grabbed his heavy, hard-leather briefcase, clenched his pipe between his teeth, and strode through the halls of the Pentagon to his office in the D-ring, stepping neatly around the soldiers on adult-sized tricycles who rode up and down the endless corridors and the ramps between floors, delivering mail. Or he would skip the office and head to the airport for a trip to Boston or Pittsburgh or Palo Alto, so he could check on the research he was funding with an annual budget of $15 million to advance computer technology.
Taylor and Wessler worked at the Advanced Research Projects Agency (ARPA), which oversaw research initiatives for the Department of Defense. (ARPA is today called DARPA.) Taylor ran ARPA’s Information Processing Techniques Office. The $15 million he had to work with was only a tiny fraction of ARPA’s $250 million annual budget, but he could direct it all toward computing.3
The Department of Defense believed that advanced computing would give the United States an edge over the USSR, and Taylor, too, was an ardent believer in the promise of computers. He thought that no other technological field was “able to so strongly affect all other areas of human endeavor.” Moreover, he was confident that “in no other domain is there a greater technological distance between the United States and the rest of the world.” Computers, he believed, would help the United States win the Cold War.4
Walking with his protégé, Taylor stopped for no apparent reason. “Let’s take a little detour, Barry.” Taylor tended to bark, not because he was angry but because he was impatient. He pivoted on his heel and began walking in the other direction, Wessler close behind.
A few minutes later they were beelining through the E-ring, where the muckety-mucks of the Pentagon had offices behind thick wooden doors guarded by secretaries. With a jovial “Hi. I’m just going in to talk to him,” Taylor swept past the secretary of an assistant secretary of defense. Before the desk attendant could even push back her chair, Taylor had opened the office door, slipped inside, and taken a seat.
Wessler sat down next to him, stunned. Only minutes before, Taylor had been complaining about the byzantine rules of communication within the Pentagon, laying out how if one man wanted to speak with another of higher rank or status, protocol dictated that the lower man’s secretary call the higher one’s secretary to set up a phone call, and then, at the agreed-upon time, the lower-level man’s secretary had to call the higher man’s secretary and then get the lower-level man on the line before the higher-level man’s secretary would bring her boss on. “A lot of work just to talk to a guy,” he had groused.
Barging into the office of a man several reporting levels above him, Taylor had just broken every rule.
Meanwhile, the assistant secretary, surprised at his desk, had risen to his feet, face blazing. His secretary began apologizing.
Taylor began talking about the projects his office was funding: what he was excited about, where he had concerns. The next thing Wessler knew, the assistant secretary was walking back around to his seat, dismissing his secretary, listening, and nodding his head.5
The primary function of bureaucracy, Taylor believed, was to block communication, and nothing irritated him more than obstacles that prevented people from sharing ideas and getting feedback. Taylor, like any good Texan, loved football—he always regretted having been too small to play in college—and for his entire career he anticipated any encounter with a bureaucratic system as a new play to strategize. Should he barrel through, as he did in the assistant secretary’s office? Or should he finesse the situation to his advantage? Both finesse and assertiveness could get him what he wanted.
It was Taylor’s good fortune that the director of ARPA, Charles Herzfeld, felt the same way about the Pentagon’s bureaucracy. Herzfeld, a University of Chicago–trained physicist from Vienna, was, in one memorable description, “an old Washington hand who knew high-level B.S. when he heard it and wasn’t afraid to call it by name.”6 An informal conversation between the two men—one that came about when Taylor raced up to Herzfeld’s office in the E-ring and rapped on the door—led to the October 29 transmission that brought the Arpanet online.
When Herzfeld invited him in, Taylor saw no need for a preamble.
“I want to build a network.”
Tell me more, Herzfeld responded.
Bob Taylor had taken an unconventional path to become what he liked to call a “computerist.” His first encounter with a computer had been a nonencounter. Taylor was finishing a master’s thesis in physiological psychology at the University of Texas. His research, to determine how the brain locates the origin of a sound, had yielded pages of data that he needed to analyze. His adviser sent him to the university’s computer center to run the data through a statistical tool on the school’s prized machine.7
The computer center was cavernous. A wall separated the room-sized computer from its users. In front of the wall, a student in a white lab coat sat at a high desk. Taylor could not see much of the machine, but he could hear its dull humming behind the young sentry.
Yes, the student said, Taylor could use the computer. He went on to explain how. First, Taylor needed to take an empty stack of cards that were roughly the size of dollar bills to one of the center’s keypunch machines. He then needed to type in his data by punching a specific pattern of holes, up to 72 digits per card, that corresponded to his data set. The computer would read the cards—
Taylor interrupted, not sure he had heard correctly. “You mean I have to sit down and punch holes in these cards to get my data in, and then I have to take the cards over to the computer, and I give the cards to a guy who runs them through the computer, and I go away and come back and get the results on a long printout of paper?”
The student said that Taylor had it right. It could take several hours or even days, depending on demand, for the printout to be completed.8
Taylor thought that system was ridiculous. “I’m not going to do that,” he said.
He left the building, irritated. There’s got to be a better way, he thought. It would be just as fast to input by hand every calculation into the lab’s Monroe calculator, a heavy, typewriter-sized beast with more than a hundred round clacking keys. And that was exactly what he did.
About a year after graduating, Taylor took a job at the Glenn L. Martin Company (now Martin Marietta) as an engineer to work on the system design for the Pershing missile. It was a desk job with a twist. He analyzed and integrated the computerized test and checkout processes, and he also occasionally donned a special suit and ran parts of the missile through trials in a cold-test lab so frigid that he was warned to blink frequently so his eyeballs did not freeze.9
During a break at Martin, Taylor read a paper that changed the course of his career. Just over seven pages long, it appeared in a new technical journal, IRE Transactions on Human Factors in Electronics. The paper addressed what its author, J.C.R. “Lick” Licklider, called “Man-Computer Symbiosis.” Like Taylor, Licklider was a quantitatively oriented psychologist whose work was more about the physics of perception than the nuances of private feelings.
The paper was less a report than a vision: “The hope is that, in not too many years, human brains and computing machines will be coupled together very tightly, and that the resulting partnership will think as no human brain has ever thought and process data in a way not approached by the information-handling machines we know today.” The computer, in Licklider’s imagining, could become something more than a tool for calculation; if responsive and easy enough to use, it could help a person to think and create.10
“When I read [the article], I just lit up,” Taylor recalls. “I said, ‘Yes. That’s the answer to my key punch dilemma. That’s worth working on.’ ”11 For years, he had been hearing that a computer was like a giant brain, but he had never before understood the comparison. The overgrown calculator in the University of Texas computer center and its cousin machines around the world were nothing at all like the human brain. Licklider’s interactive computers, however, were brainlike: tools that extended humans’ capacities to think and create.
Two years after reading Licklider’s article, Taylor met the author. Taylor was then working at NASA, allocating money for computing research, and Licklider was running the Information Processing Techniques Office at ARPA, a job Taylor would himself assume a few years later. The two men became friends, so much so that Licklider’s wife, Louise, would later tell Taylor, “You are a very special son to us, and we love you.”12 Licklider and Taylor traveled together for work and enjoyed many evenings on the road in small, unassuming bars. Sometimes they would talk about their jobs, sharing ideas about the researchers who were pushing to attain Licklider’s vision, but just as often they shared personal stories. One memorable time, Taylor, who sang and played guitar, took the stage with a folk band in Greece, and Licklider insisted on staying until early morning to hear every piece his colleague played.
Taylor joined ARPA’s Information Processing Techniques Office in 1965 as assistant director to Licklider’s successor, Ivan Sutherland, another leading proponent of interactive computing. Taylor and Sutherland worked closely together—they were the only people in the office, aside from a secretary—and Taylor enjoyed traveling to the many campuses where ARPA had funded the development of new computer science programs. Before 1960, there were no computer science departments. People interested in computing clustered in engineering, math, or physics departments. But the ARPA money allocated by Licklider, Sutherland, and Taylor had changed all that. Now schools such as MIT, Carnegie Mellon, the University of Utah, the University of Illinois, the University of California at Berkeley, and Stanford had fledgling computer science departments. Many of these departments were built around a new type of computer whose development ARPA had also helped fund. The machines, called time-sharing computers, could switch from running one person’s program to running the next person’s so quickly that each user, sitting alone at one of many typewriterlike terminals, often had the illusion that the computer was working exclusively on his or her program. Time-sharing computers did not require punch cards and displayed some results immediately.
Taylor, who racked up more than 500,000 miles on United Airlines alone during his four years at ARPA, developed a bit of a routine for each visit to the universities’ computer science programs.13 He would rent a cheap car, head to campus, and, after following up with the research team to offer assistance, would find the graduate students who cared about computing. Graduate students, Taylor felt, did the most interesting work in the field.
To Taylor’s surprise, many students told him that the time-sharing computers had helped them find friends and colleagues. A person could log on to the computer, look in a directory to see what had changed since his or her last login, and then, if a new program or clever bit of code had been added, it was easy to tell who was responsible and invite the coder for a beer. Licklider had envisioned computers that helped people to think and create by connecting one person to one machine. Now Taylor saw the potential for computers to do even more: to connect one person to another. A time-sharing computer, in one sense, was the hub of a wheel with people at the end of each spoke.
Taylor believed in the power of communities built around a common core. A clinical psychologist (of the sort Taylor did not want to be confused with) might point out that Taylor had been raised in precisely this type of community.IV In 1932, when Taylor was twenty-eight days old, a Methodist minister and his wife had adopted him. Taylor describes his father as a quiet intellectual (“not the pulpit-pounding preacher you might imagine”) and claims that he “introduced southwest Texas Methodism to existentialism in the late 1940s.” Jobs were scarce in Texas in the middle of the Depression, and Taylor’s earliest memory, from when he was three, is leaving church with his parents after a prayer meeting and finding the back seat of the family car filled with food—his father’s pay for the month.
By the time Taylor was twelve, he had lived in six towns, most of them tiny. Yet wherever he lived, he was immersed in a world of church socials, church potlucks, church services, and church camp. He spent several summers at a camp in the Texas hill country set aside for Methodist ministers and their families. It is easy to imagine that from his peripatetic-but-rooted childhood he imbibed a lesson that would help him understand the potential of a network connecting people around the world: that community comes not from geography but from shared beliefs and interests.III
Taylor was particularly interested in connecting the right people with one another. For a minister’s son who had once wanted to be a minister himself, he could be merciless in his judgments. To him, people were either geniuses or duds. He was interested in the geniuses. Once he found someone worth finding, he considered it his calling to help that person meet and work with others of the same caliber.
To this end, one of Taylor’s early moves at ARPA was to revamp the mandatory meetings of the lab leaders, or “principal investigators,” funded by the Information Processing Techniques Office. Taylor turned the principal investigators’ meetings into destination events. Under Licklider and Sutherland, the meetings had been relatively short—a few hours, maybe—and folded into a larger meeting such as the Fall or Spring Joint Computer Conference, which the principal investigators would be attending anyway. Now Taylor scheduled them for appealing times and places (Alta, Utah, during ski season; New Orleans during Mardi Gras; Hawaii in the dead of winter) and planned the formal sessions so as not to conflict with the best hours on the slopes or watching the parades in Fat City. Taylor wanted the principal investigators to interact again and again, not only around conference tables but also on a chairlift or at the beach.
At the formal sessions, Taylor asked each principal investigator to give a talk, urging him to prepare the kind of presentation he would like to hear. After the talks, he says, “I got them to argue with one another, which was very healthy, I think, and helpful to me because I would get insights about strengths and weaknesses.”14
In the meetings, Taylor set himself up as an outsider, more conductor than musician. For him, it was not a difficult role to assume. Nearly everyone else had a PhD in physics, engineering, or math. (Computer science was still such a young field that the first generation of PhDs was only beginning to graduate.)15 Taylor, meanwhile, had a master’s degree in psychology. Almost everyone else in the room came from one of the coasts or had been educated there. Not Taylor. In a Texas twang, he talked up his summer jobs herding cattle—he had once climbed onto the back of a 2,000-pound Brahma bull, just to see what would happen—or working in the oil fields.
Years later, he would discover that some of the principal investigators had been unhappy to have him in charge. He was the only head of the Information Processing Techniques Office not to hold a doctorate and the only one to have taken the job without a significant record of contributions to the field.
Taylor nonetheless says that for the entire time he was at ARPA, he had no sense of inferiority, even though he knew that he had not been the top choice to succeed Ivan Sutherland as director of the office.16 “There was not a one of them who I thought was smarter than me,” he says. (“I thought Lick was smarter than me,” he adds.) Taylor had graduated from high school at sixteen, had an IQ so high (154) it was written up in his small-town newspaper when he was a child, and finished college with a double major in psychology and math, as well as substantial course work in religion, English, and philosophy.17 He did not possess the principal investigators’ depth of computer science knowledge, but the breadth of Taylor’s mind was wide and his chutzpah, limitless. He was unafraid to fire question after question in meetings, an assertive technique that nonetheless often helped the experts clarify their own thinking. (As Barry Wessler put it, “Bob would not necessarily have a firm technological grasp on the details, but you would come in because you had a problem, and you’d walk out with a solution.”) Taylor’s different background also meant he had nothing invested in any technical or social orthodoxy and could maintain a useful distance from the principal investigators. He had not gone to school with any of them. He was not their colleague or student, nor would he train students who might want to work in their labs.
As Taylor listened and probed at the principal investigator meetings and on university campuses, an idea began to pull at the edges of his mind. Almost all of the principal investigators worked on time-sharing computers; why not come up with some way to connect the computers to one another? If connecting a few terminals to a time-sharing computer had brought together diverse users on the individual campuses, imagine what could happen if each of those time-sharing machines, with its dozens of users, was connected to other time-sharing machines. The community would grow and grow, until it became what Taylor would come to call a networked “supercommunity” or “metacommunity.”18 A network, he wrote in 1968, could “overcome the geographical barriers which prohibit the cooperative working together of people of common interests—be they students, scientists, soldiers, statesmen, or all of these.”19
Taylor resolved to build this network one day when he was back at the Pentagon and happened to turn his head to look at the entrance to the small room next door. He spent a lot of time in that room, which housed three computer terminals, each one linked directly to a different time-sharing computer: one at MIT, one at Berkeley, a third at the System Development Corporation in Santa Monica. The three computers could not talk to one another.20 But Taylor, who participated in all three computing worlds, knew that the ideas and problems being addressed in each place were similar. “You don’t have to be half smart to see this thing ought to be designed such that you have just one terminal and you can go wherever you want to go,” he said years later, recalling his thoughts when he had looked at the array of machines and imagined a network connecting them.21
Taylor said that he went to ARPA wanting to do “one big thing,” and in this computer network, he saw his opportunity. Networks had come up before in a “wouldn’t it be nice” kind of way, and Licklider had outlined his thoughts on what he called an “intergalactic network” in 1963 in a memo to principal investigators. By the fall of 1966, when Taylor turned to look at the three terminals, he and Licklider had been friends for years, but Taylor says he had never connected his ideas to Licklider’s imagined “intergalactic network.” Nor did he and Licklider talk about a computer network until after the Arpanet project was under way, he said, though Licklider supported the idea once he heard it. Taylor says, “It could be that our discussions subconsciously prompted me to say maybe we can connect all these computers. I just know that it wasn’t explicit.”22
Taylor moved the network idea from “wouldn’t it be nice” to “we can do this—now.” He understood that there were enough people in computer science—hundreds—who would value being networked to each other and who also could figure out how to build the network. To demonstrate that a network was technically doable, he funded a small experiment to send a few bits back and forth, rather slowly and across telephone lines, between a TX-2 computer at MIT’s Lincoln Laboratory and an incompatible Q-32 computer at System Development Corporation. The experiment also made clear that long-distance communication between two different computer systems, though feasible, would require a huge amount of work to be made quick, reliable, and scalable.
There was one more reason why the time to build a network was now, Taylor knew: ARPA would pay for it. If the project had to be funded piecemeal, with different government agencies or universities each financing a portion of its development, the work would be slowed down by endless bickering around fair shares and conflicting reporting procedures and requirements.
Taylor likely marshaled all these arguments in the fall of 1966 as he sat forward on his chair in Charles Herzfeld’s office, explaining why ARPA should fund a computer network. (Neither Taylor nor Herzfeld has an exact memory of the conversation, nor is there a written record.) Taylor’s primary interest was in connecting ARPA-funded researchers and thereby amplifying the power of their brains. He told Herzfeld that a computer network would enable principal investigators at one site to use programs and data at other sites, thereby cutting down on wasteful duplication. Moreover, because the network could connect different types of machines, there could also be a direct payoff for the Department of Defense: a network would free up people in the field or at the Pentagon to buy the best computer for a given job, rather than being locked into a single system for everyone. And, Taylor continued, it was worth building this network only if it could be fast. Otherwise, “the illusion of being a local user, near the computer, [would] be destroyed.”23 Contrary to popular beliefs about the origins of the Internet, there is one important argument that Taylor did not make for the network: he never claimed that the network would be good for national security.24 Later, Herzfeld and his successors, petitioning Congress to continue funding for the Arpanet, would argue that the network could enable communications to continue even in the event of a nuclear attack. But that was not a motivation for the network’s creation.
Herzfeld had been quiet for almost half an hour while Taylor made his case. Both men knew that the Information Processing Techniques Office budget was committed for the fiscal year. But it was not unusual to fund a new project using money earmarked for something else and then to write the new project into the next year’s budget as maintenance of an ongoing effort. Congress seemed much more inclined to fund projects already under way than to take risks on new ones.
Taylor was raising his eyebrows, as he often did to indicate that it was someone else’s turn to talk.
Herzfeld nodded. This was not the first time he had heard of a network. Years earlier, he had listened as Licklider had shared his vision of a giant computer network. It had been just a dream at that stage (as its “intergalactic” name made clear), but now, according to Taylor, the network could be real.25
“Do it.” Herzfeld said. He told Taylor he could spend at least $500,000.26
Convincing other people was much more difficult. The network project was a departure for both the Information Processing Techniques Office and ARPA. Under standard procedure, a researcher, usually at a university, would initiate a project and submit a proposal for funding. Often the “proposals” were conversations, since most of the researchers were known to Taylor. Indeed, many formal proposals were written months after the projects had been funded.
The network project, by contrast, was initiated by ARPA—more specifically, by Bob Taylor, who was determined to see the network succeed. Taylor’s first move was to hire Al Blue, an expert in ARPA’s legal, contractual, and financial protocols, to document the network program. Blue, who had worked at ARPA for years, could not believe the “strange dichotomy” between the computing world surrounding the Information Processing Techniques Office and the reality experienced by the rest of ARPA. While Taylor’s office was funding state-of-the-art time-sharing machines and now a new computer network, the rest of ARPA ran on the punch-card batch-processing computers of the sort that Taylor had refused to work with when finishing his thesis. For certain reports, Blue would leave his Information Processing Techniques Office with its sleek terminals and high-level talk about coding and networking to drop off his hand-annotated papers for an operator at Cameron Station to keypunch onto cards for the computer. With his new perspective from Taylor’s office, Blue now saw the approach he had been using for years as “primitive.”27
It is a measure of how much the network project meant to Taylor that, to ensure its launch, he was willing to increase his bureaucratic and paperwork responsibilities to the point that he needed to hire new staff. In addition to Blue, he needed a program manager. The network, only one of seventeen projects Taylor was funding, was going to send messages among incompatible computing systems at a time when there were no agreed-upon standards for operating systems, programming languages, or word sizes.28
Taylor had neither the technical expertise nor the time to take day-to-day responsibility for the network’s construction. He needed somebody whose technical instincts he could trust and, moreover, someone whom the principal investigators could respect. He knew whom he wanted for the job: Lawrence G. Roberts, a lead researcher at MIT’s Lincoln Lab. Roberts held a PhD in engineering from MIT and had made major contributions in the field of computer graphics before he worked on the small proof-of-concept networking project that had brought his talents to Taylor’s attention. He was a tenacious and productive scientist, and already, at twenty-nine, a leading figure in the field.
There was only one problem. Roberts did not want to run the networking project. He did not want to leave his own research. He did not like the idea of reporting to Bob Taylor, not because of anything personal about the man, but because in the universe in which Larry Roberts was educated and now lived, it just wasn’t right for a guy with an MIT PhD in engineering to report to a guy with a master’s degree in psychology from the University of Texas.
So when Taylor asked Roberts to come to ARPA to take charge of a new computer network, Roberts said no. He refused again a few months later and again after that.
But Taylor wanted the network to happen, and he wanted Roberts to make it happen, and Taylor was stubborn. He found a way to get Roberts—blackmail.29
Taylor asked ARPA director Charles Herzfeld to call MIT’s Lincoln Lab and remind the head of the lab that 51 percent of its funding came from ARPA. Then he asked Herzfeld to suggest that the head of the lab let Roberts know that it would be in his best interest, and the best interest of the lab, to take the ARPA job overseeing the network.
Not too many days later, Roberts was called into the office belonging to the head of Lincoln Lab and told, “It’d probably be a nice thing for all of us if you’d consider this.”30
By December, Roberts was in Washington, DC. The next month, he reported to his new job as program manager for the network.31
Privately, however, Roberts did not think of himself as the program manager. Instead, he was the office director in waiting, who would assume control as soon as he had a chance to learn the paperwork and reporting niceties of the job. Taylor was a placeholder for the succession that should have happened, a succession that Taylor’s predecessor Sutherland had wanted: from Licklider, to Sutherland, to Roberts.32
It is a credit to both Taylor and Roberts that their differences never surfaced publicly. When the Roberts family first moved down to DC from Boston, they spent the Christmas holidays with Taylor’s family. Within a short time, such socializing ended, but Taylor and Roberts always maintained a cordiality in the office, even after Roberts was named chief scientist and began reporting to the deputy director of ARPA, rather than to Taylor; even after, in some people’s minds, Roberts seemed to want people to think that he, not Taylor, ran the Information Processing Techniques Office.
Taylor let Roberts say what he wanted. What mattered was getting the network built, and Roberts was the man to do it. Hiring him, Taylor later said, “was probably the hardest and most important [element] of my Arpanet involvement.”33
Roberts, five years younger than Taylor, was slight and tall, with a cleft chin, intense dark eyes, and a hairline that appeared to have been beaten into retreat by his blazing mind. Roberts ran at only two speeds: off and full throttle—and no one ever saw the first. “When you were working with Larry, you came into the office and you hit the ground running, and you did not quit running until the whistle blew, and then you ran some more. He was on the go every minute. When Larry ate a meal it was like shoveling coal into the steamboiler,” Blue recalls.34 He proved an inspired choice to run the network.
In March 1967, Roberts and Taylor, jointly leading a meeting of ARPA’s principal investigators in Ann Arbor, Michigan, told the researchers that ARPA was going to build a computer network and they were all expected to connect to it. The principal investigators were not enthusiastic. They were busy running their labs and doing their own work. They saw no real reason to add this network to their responsibilities. Researchers with more powerful computers worried that those with less computing power would use the network to commandeer precious computing cycles. “If I could not get some ARPA-funded participants involved in a commitment to a purpose higher than ‘Who is going to steal the next ten percent of my memory cycles?,’ there would be no network,” Taylor later wrote. Roberts agrees: “They wanted to buy their own machines and hide in the corner.”
In the face of this lack of interest, Taylor and Roberts increased the pressure. They explained that ARPA would not fund new computers for principal investigators until the existing computers were networked and, in Roberts’s words, “you have used up all of the resources of the network.” Taylor and Roberts also said that if the principal investigators wanted to make sure the network did not become the resource drain they feared, they should help design it. 35
The tenor of the meeting began to change as many of the principal investigators, despite their initial reaction, became intrigued with the concept of a network. Though trepidatious, they could not resist the appeal of the problems involved. As Robert Kahn, an engineer who joined the project later, put it, “The effort to actually create the ARPANET design was actually a pretty intense intellectual activity . . . like the first missile that you send up, or the first rocket into space; you don’t know if it’s going to work or survive.”36 The opportunity to build something from scratch, using cutting-edge ideas and designing every part of the system, appealed to the researchers. Roberts led a technical discussion that grew quite animated, while Taylor kept pace, making sure that the fine-sounding ideas could be translated into practical reality. “Somebody would say, ‘Well, that is going to mean that you will have to buy a bunch of something-or-others,’ ” recalled MIT’s Wes Clark, who was at the meeting, “And Bob Taylor would say, ‘Well, let’s see, I guess we can just take that out of something-or-other program funds.’ And he would turn to Al Blue [the expert in ARPA compliance] and say, ‘Can we do that, Al?’ And Al would think for a half a second and say, ‘Yes, we can do that.’ ”37
The technical discussions in Ann Arbor ran so long that they continued while Taylor drove a number of participants to the airport. During the car ride, Wes Clark proposed an idea: rather than networking computers together directly, the group should build an underlying backbone network of routers to streamline connections between computers and help make the network more reliable.38 His contribution was critical.
Taylor kick-started the network. He got it funded. He helped to cajole researchers to contribute to it. No wonder Paul Baran, the inventor of packet switching, called the Arpanet Taylor’s “baby.”39 But after the Ann Arbor meeting, once it was clear the network was going to be built, Taylor stepped back and Roberts stepped forward.
Roberts spent hours sketching out possible network configurations and meeting with principal investigators in subgroups that addressed topics ranging from software protocols to hardware design, to appropriate bandwidth allocations, to packet storage and routing. By summer 1967, a number of key issues had been resolved. The network would be packet-switched and run on a subnet system of the sort Clark proposed. The first four nodes would be at UCLA; Stanford Research Institute; the University of California, Santa Barbara; and the University of Utah. Consensus usually dictated such decisions, but it was Roberts who cast the deciding technical vote if necessary.
By mid-1967, Taylor’s primary role in the network project had shifted to keeping his bosses at ARPA informed and excited so that funds would flow unimpeded. He spent much of the Christmas holidays at Licklider’s house, coauthoring an article to introduce the notion of networking to the business world. “In a few years, men will be able to communicate more effectively through a machine than face to face,” Taylor and Licklider wrote in “The Computer as a Communication Device,” published in Science and Technology the following spring.40 The coauthors explained how networks would enable people to “interact with the richness of living information” in a new way, not as passive readers or learners but as active participants. The article ended with a prophetic statement about networking: “For the society, the impact will be good or bad, depending mainly on the question: Will ‘to be on line’ be a privilege or a right? If only a favored segment of the population gets a chance to enjoy the advantage of ‘intelligence amplification,’ the network may exaggerate the discontinuity in the spectrum of intellectual opportunity.” Taylor estimates that this article was seen by “essentially no one.” The magazine (“for the technical men in management”) failed soon after the article was published.II
Taylor says that he intervened in the network project one other time after the Ann Arbor meeting. In July 1968, the Information Processing Techniques Office issued a request for bids to build the network. This huge job required writing complex software, designing and building specialized routers called interface message processors (IMPs), and any number of other tasks. Twelve companies submitted bids.41 Taylor says that a subcommittee of the networking group selected a large military defense contractor, Raytheon, to do the job. Taylor thought Raytheon was a terrible choice. The winning company would have to work with the academic researchers. Raytheon’s command-and-control culture would not mesh well with the “you can’t make me do it” academic culture. “It was just oil and water; it would not mix,” he says.42
Taylor overruled the subcommittee and awarded the contract instead to Bolt Beranek and Newman, a small firm where Licklider had once worked and that was staffed mostly by MIT grads. Taylor believed the cultural fit would be better.43 The issue was not technical, but about human interactions, and the person with the final say was not an engineer but a psychologist.44
Taylor had many more responsibilities than the network, which by this time was being called the Arpanet. He was involved in the launch of a new program in computer graphics at the University of Utah, a program whose graduates would come to include the founders or key minds behind Pixar-Disney Animations, Adobe, Silicon Graphics, Xerox PARC, and Atari.45 Forever concerned about getting like-minded people together, he also initiated a conference for graduate students working with his principal investigators. Many of those students would go on to do pioneering and critical work in computer science.
Taylor continued to travel to visit principal investigators, all of whom had ARPA-funded projects they were working on in addition to the network. A demonstration by one of those researchers, a craggy-faced forty-two-year-old engineer named Doug Engelbart at the Stanford Research Institute, impressed Taylor. Taylor had quietly been supporting Engelbart’s rather out-there research since 1961, when Taylor, then at NASA, had funded Engelbart to build a new interactive computing system that included, among its many innovative features, the world’s first computer mouse.46 (Taylor later said, “Remember when NASA was advertising Tang [the bright-orange-colored drink consumed by astronauts] as its big contribution to the civilized world? Well, there was a better example, but they didn’t know about it.”) Taylor’s respect for Engelbart only grew when, in the first meeting about the Arpanet, the researcher, alone among the principal investigators, expressed immediate support. Engelbart volunteered Stanford Research Institute to archive documents related to the project and make them available as paper copies or, later, via the network itself.
When he had moved to ARPA, Taylor had continued to fund Engelbart’s work, directing some $500,000 (nearly $3.6 million in 2016 dollars) to the researcher. He was forever pushing Engelbart, who was by nature soft-spoken and self-effacing. At a dinner in about 1966, Taylor turned to the researcher abruptly. “The trouble with you, Doug, is that you don’t think big enough,” he said. “What would you really want to do?” When Engelbart, flustered, admitted that if he had enough money, he would buy a million-dollar time-sharing computer that would enable him to test and evolve his ideas more quickly, Taylor said, “Well, let’s write a proposal.” And they did.47
In October 1967, Taylor went to the Stanford Research Institute to attend an invitation-only review of Engelbart’s new computer system, called NLS (for oNLine System). Engelbart assembled the few visitors at a number of terminals, and, wielding his newfangled mouse in one hand and a specialized chord keyboard he had developed in the other, he demonstrated the closest approximation of Licklider and Taylor’s vision for computing that Taylor had seen. Here was Engelbart, showing how he could edit documents by pointing and clicking on a screen, in 1967, when most machines could only show information by spitting out fanfold paper. Here was Engelbart toggling between multiple windows. Here was Engelbart urging the participants to play with their terminals and see how they could pull up new information by highlighting certain words: the world’s first demonstration of hyperlinks. Here was Engelbart giving what amounted to an interactive rudimentary PowerPoint presentation. Anything he typed on his computer would appear on the other terminals, and if anyone at those terminals moved a mouse to track a spot on his individual screen, the tracking was visible on all the other terminals.
Taylor had been promoting interactive computing for years, and here was Engelbart doing it. The presentation was not a complete surprise, since Engelbart had kept Taylor abreast of his progress, but it was impressive nonetheless.
There was even more to come. Engelbart explained that everything he was doing could work on the Arpanet, once the network technology was sophisticated enough. A researcher in California would be able to point to a spot on his interactive terminal, and thousands of miles away, at MIT for example, someone else on a networked computer would be able to see that cursor move.
Taylor told Licklider about the demonstration, and the two men included a detailed description of it in their 1968 “The Computer as a Communication Device” article. They told their readers, “If our extrapolation from Doug Engelbart’s meeting proves correct, you will spend much more time in computer-facilitated teleconferences and much less en route to meetings.”
More significantly, Taylor insisted that Engelbart find a way to show more people what this system could do. “I was working with the rest of the computer world, as well as with Engelbart,” he recalls. “Engelbart’s project was way off in weird land, compared to all my other projects, and I wanted the other project people to see it.”
“This was the future of computing, as far as I was concerned,” he explains.48 While the field of human-computer interaction was attracting more interest every year, only a handful of people—Engelbart’s lab team and the invited participants in the project review—had seen a physical demonstration of its possibilities, and then only in a low-key way. Reorienting the trajectory of computer science would require something much bigger and much more dramatic.
You need to show more people what you are doing, Taylor told Engelbart again at the end of the invitation-only demo. Engelbart admitted he had considered it, largely because he was afraid that “if we didn’t get people’s attention, ARPA would have a hard time maintaining its support for us.”49 Taylor and Engelbart batted around ideas of venues that might make sense for a public demonstration and settled on the Fall Joint Computer Conference. The conference, one of the young industry’s two most important, would be held in San Francisco in December 1968. More than a thousand people, many of them researchers, would attend.
Engelbart liked the proposed venue but still had objections. A demonstration would be expensive. Since he could not move his system, which ran on the giant time-sharing computer Taylor had helped him write the grant for, the only way to do a demonstration would be to set up a huge display in San Francisco and somehow connect it to the Stanford Research Institute machine room some thirty-five miles south.
Taylor would not be deterred; he said his office would pay for the demonstration.50 In making that offer, he was departing from his brief at ARPA. Instead of backing advanced research, he was supporting evangelism.I But he pushed ahead, convinced that a public demonstration of Engelbart’s work would advance the field. He helped in other ways, as well. When Engelbart’s lead engineer, Bill English, determined that the only projector that could give the right effect belonged to NASA, Taylor vouched for the lab.
Engelbart’s December 9, 1968, demonstration at the Fall Joint Computer Conference has come to be known as the Mother of All Demos. Even the world’s most advanced computer scientists had never encountered anything like his system.
What a show it was. The San Francisco auditorium seated two thousand, and Engelbart’s presentation had been given the full ninety-minute slot typically split among a panel of presenters. Engelbart sat on the right side of the stage, facing the audience, his headset resting on the silvering hair at his left temple, his hands hovering above an unusual workstation that housed a standard keyboard, a special five-key chord key set, and the mouse that Taylor had seen at the invitation-only event. Above and behind Engelbart hung a twenty-two-foot-high screen that alternately projected the image of his face and the display output of his workstation’s monitor—and sometimes both at the same time, side by side or one overlaid on the other.
When the lights went down, Engelbart began speaking. His voice, amplified through the loudspeaker from the microphone on his headset, was confident, if a bit halting. “If in your office you as an intellectual worker were supplied with a computer display backed up by a computer that was alive for you all day and was instantly responsible . . .” Prompted by Bill English, who was stage-managing the demonstration and speaking straight into Engelbart’s headset, Engelbart corrected himself: “responsive to every action, how much value could you derive from that?” He allowed himself a small smile. “I hope you’ll go along with this rather unusual setting. If every one of us does our job well, it will all go very interesting.” He then whispered to himself: “I think.”
For the next ninety minutes, Engelbart showed off the same technology that had captivated Taylor. But here, so much bigger, so much louder, with the audience’s murmurs of surprise and approval punctuating the demonstration, Engelbart’s system to “augment human intellect” seemed even more incredible. Throughout the presentation, his computer buzzed and beeped, which only contributed to the overwhelming sense that this was some sort of science fiction extravaganza from a futuristic research pod.
Engelbart pulled up an imaginary shopping list and showed how he could reorder and reorganize it with the click of a mouse. He opened a link to a map. About halfway through the presentation, he explained that he wanted to connect to his lab team at the Stanford Research Institute in Menlo Park. He had been remotely using the computer at the lab, an impressive feat in itself, and now he wanted to bring in video images.
“Come in, Menlo Park,” he said. The entire audience could hear him inhale and hold his breath. Engelbart knew that connecting with Menlo Park required the perfect synchronization of two custom-built modem lines and two video microwave links relayed to San Francisco from a truck parked on a hilltop midway between the city and the laboratory.
Only after a new image flickered onto the screen—a young man’s well-manicured right hand, grasping a mouse, hove into view—did Engelbart exhale and resume his talk: “Okay, there’s Don Andrew’s hand in Menlo Park.”
And he was off again. He introduced the mouse (“I don’t know why we call it a mouse. Sometimes I apologize. It started that way, and we never did change it.”). He showed the hardware that was driving the system. He demonstrated how someone in Menlo Park could see the same document that Engelbart had on his screen and how, if the man in Menlo Park moved his mouse, the cursor (Engelbart called it a “tracking spot” or “bug”) moved on Engelbart’s own screen projected for the auditorium.
Engelbart also gave a nod to the Arpanet. Noting that his lab would be the second node on the network and the keeper of the network’s library, he explained that the network designers “plan to be able to transmit across the country [fast enough] that I could be running a system in Cambridge over the network and getting this same kind of response.”
As the presentation neared its end, Engelbart began thanking people. He mentioned his seventeen-person team. He dedicated the show to his wife and children, who were in attendance. He singled out Bill English, who had prepared and stage-managed the entire “performance,” as Engelbart called it.
Engelbart mentioned only one other person by name. He thanked Bob Taylor “for backing me all these years in this wild dream of doing this sort of thing.”
The demonstration ended. The clapping was loud and sustained, and after a few seconds, the audience rose to its feet for a standing ovation. Watching Engelbart “dealing lightning with both hands,” in one memorable description, counted for many as a near-religious experience.51 Taylor says with satisfaction that “Nobody’d seen anything like it.”
In the spring of 1969, Taylor received good news about the computer network. The contractor Bolt Beranek and Newman, the firm Taylor had chosen over Raytheon, had used the IMP router network to transmit data between two computers acting as nodes. Taylor saw this feat as solid evidence that the network was going to work—perhaps not instantly (he knew bugs would need to be worked through) but eventually. He had accomplished his goal of doing one big thing during his ARPA tenure.
And now, he felt, he was ready to move on. He had been at ARPA for five years, longer than average.52 He had an excellent successor in Roberts.
He was ready to leave for another reason, as well. As the war in Vietnam escalated, he no longer wanted to be associated with the Department of Defense. In the first stages of the conflict, he, like many of his peers, had been moderately supportive: the South Vietnamese needed aid, he thought. So in 1967, when he was asked to go to Vietnam to help reconcile conflicting reports from computers at various bases, he had reported to an office in Virginia, where he was given a cholera shot and an ID card listing his rank as a general in the US Army, a rank he enjoyed both for its absurdity (he had last fired a gun in high school) and for its status.53 He was given two copies of the card; one, he was told, was to give to the enemy in the event of his capture.
A few weeks later, Taylor was bound for Vietnam with three friends from the office of the Joint Chiefs of Staff. The trip got off to an auspicious start, with an overnight in Hawaii. When the plane stopped on the tarmac, a driver pulled up in a pristine car, jogged to the stairs that had been pushed up to the plane, and unrolled a red carpet from the bottom step right to the rear door of the automobile.
One of Taylor’s traveling companions nudged him. That’s for you.
“What is this for?” Taylor asked.
“You’re a general. These people are assigned to you, to look after you while you’re here.”
Taylor laughed. “I can’t do this. This is just silly.” He asked one of the men from the Joint Chiefs office to send the car away. The carpet was rolled up, the car drove off, and the four friends rented a cheap clunker so they could, in Taylor’s words, “do crazy things” that he declined to specify.54
Taylor’s early days in Vietnam seemed to confirm his belief in the value of the war. South Vietnam was breathtaking. The women were lovely—one, in fact, kept trying to convince Taylor’s traveling companion to arrange for her to sleep with the handsome young general.55 After days visiting the bases, Taylor would return to his hotel room and shake his head at how once beautiful Saigon was now buried beneath sandbags and military vehicles.
He soon grew disillusioned. People who had been in the country for a while told him that the South Vietnamese government ran on patronage and nepotism; any victory by the South Vietnamese and their US allies would only install a corrupt bureaucracy. Taylor began to question the value of the war: Why fight for years and years in what is essentially a civil war, lose many Vietnamese and American lives, and then, when it’s over, leave behind a government that will not help the South Vietnamese people? The trip to Saigon marked the beginning of the end of Taylor’s time in the Department of Defense.
Opposition to the US involvement in Vietnam also led a few principal investigators to wonder if they should continue to accept ARPA money. Nearly all did. There was a general sense among researchers that Taylor’s office was more part of the research community than part of the Department of Defense monolith. “How can that bunch of guys be really interested in murdering civilians?” the researchers asked themselves.56 To be sure, by the end of the 1960s, every program funded by Taylor’s office needed to include a relevance statement addressing how the program would benefit the military, but people in the research community understood that Taylor’s office was “inventing stories” on this front. The statements were “fiction in many cases,” according to Al Blue.57
Taylor nonetheless felt that as long as he was working for ARPA, he was supporting what was happening in Vietnam. He knew that his office had paid to develop a prototype time-sharing system for the National Military Command System Support Center and that once that system was in place it would be, as the Arpanet program plan put it, “a natural recipient for an interactive computer network.”58 In mid-1968, Taylor asked Oregon Senator Wayne Morse to help him find a job outside of the Department of Defense. Morse was one of only two senators to oppose the Gulf of Tonkin Resolution, which authorized military action in Vietnam without a formal declaration of war.59
At about the same time, Dave Evans, a principal investigator who had established the University of Utah’s graphics center, offered Taylor a job coordinating a number of computing research projects from around the university.60 The job did not fire Taylor’s imagination, but it was in his general field of interest, offered by a friend, and located well outside of the Pentagon. (And, Taylor adds, it would give him a chance to work on his skiing.) The University of Utah had an outstanding computer science program, Taylor knew. Moreover, the school was slated to come online as the fourth Arpanet node in a few months.
The West itself was also a big draw. The light reminded Taylor of Texas. The air was clear. And it seemed that there was an openness in the West, a kind of acceptance based on what you were doing now, rather than on where you came from. It was no accident, he thought, that the first four Arpanet nodes were going to be in California and Utah—the westerners who worked there just seemed more open to new ideas, “less hidebound,” in Taylor’s phrase, than their colleagues at East Coast institutions such as MIT.
When the L-O-G message was sent from UCLA to Engelbart’s lab at the Stanford Research Institute in October 1969, Taylor was making his way from Washington, DC, to Salt Lake City. He was taking with him across the country his wife and three children, the family station wagon and his prized Corvette, boxes of household goods, and his Rolodex listing the researchers who he felt were doing the best computer science work in the country.
He also took with him one of the ID cards identifying him as a general. A military base not too far from the university had one of the few nearby bars, and Taylor wanted to be able to buy drinks at the canteen.
Taylor had been told that the second ID card was to give to the enemy. He left that one on his desk at the Pentagon. He had decided that the administration backing the unjust war in Vietnam was, by definition, his enemy.
Bob Taylor headed west.
I. Taylor recalls, “The regional contracting officer, who looked after DOD [Department of Defense] money in this region, saw these expenses going across his desk for these strange things, and he got in touch with Engelbart about this: ‘What’s all this about?’ Engelbart explained it to him, and this contracting officer . . . said, ‘This thing sure sounds crazy to me, and I’ll tell you what. If it doesn’t work, I’m going to deny I ever heard about it.’ I didn’t know that story until the demo was already finished, and Engelbart told me what this guy had said.”
II. The article also described an artificial-intelligence agent called OLIVER that would act as an über-personal assistant in the age of networked machines (“Lick’s idea,” Taylor says). The piece ended with a vision of nerdutopia in which all unemployment has disappeared from the planet because “the entire population of the world is caught up in an infinite crescendo of on-line interactive debugging.”
III. Taylor claims his childhood influenced him only insofar as his parents always told him he was special because they had chosen him to be their son, while most parents just had to take whoever showed up. That notion, he admits, instilled in him a certain unshakable self-confidence. Apple cofounder Steve Jobs, also adopted, said that his parents had told him the same thing.
IV. Taylor says that he did not go on for a PhD in psychology in part because because he would have had to qualify in areas such as child psychology and other “soft” disciplines that he thought were “crap.”