Where Wizards Stay Up Late (5 page)

The principal charter he'd been given was to come up with uses for computers other than as tools for numerical scientific calculations. Lick developed new programs partly as a reaction against some of the applications the Defense Department had in mind for large computers. Air Force intelligence, for instance, wanted to harness huge mainframes to detect patterns of behavior among high-level Soviet officials. The computer would be fed intelligence information from a variety of human sources, such as hearsay from cocktail parties or observations at a May Day parade, and try to develop a best-guess scenario on what the Soviets might be up to. “The idea was that you take this powerful computer and feed it all this qualitative information, such as ‘The air force chief drank two martinis,' or ‘Khrushchev isn't reading
Pravda
on Mondays.'” recalled Ruina. “And the computer would play Sherlock Holmes and conclude that the Russians must be building an MX-72 missile or something like that.”

First Ruina, then Licklider, tried putting a stop to such “asinine kinds of things,” as Lick described the ill-conceived projects. Then Lick worked to find the country's foremost computer centers and set up research contracts with them. In short order, he had reached out to the best computer scientists of the day, from Stanford, MIT, UCLA, Berkeley, and a handful of companies, bringing them into ARPA's sphere. All told, there were about a dozen in Lick's inner circle, which Ruina called “Lick's priesthood.” In typical fashion, where his most passionate beliefs masqueraded as a bit of a joke, Licklider nicknamed it the Intergalactic Computer Network.

Six months after his arrival at ARPA, Lick wrote a lengthy memo to the members of the Intergalactic Network in which he expressed his frustration over the proliferation of disparate programming languages, debugging systems, time-sharing system control languages, and documentation schemes. In making the case for an attempt at standardization, Lick discussed the hypothetical problem of a network of computers. “Consider the situation in which several different centers are netted together, each center being highly individualistic and having its own special language and its own special way of doing things,” he posited. “Is it not desirable or even necessary for all the centers to agree upon some language or, at least, upon some conventions for asking such questions as ‘What language do you speak?'At this extreme, the problem is essentially the one discussed by science-fiction writers: How do you get communications started among totally uncorrelated sapient beings?”

That said, Lick hedged his bets. “It will possibly turn out,” he continued, “that only on rare occasions do most or all of the computers in the overall system operate together in an integrated network. It seems to me to be important, nevertheless, to develop a capability for integrated network operation.” And therein lay the seed of Licklider's grandest vision yet. He would extend the concept of the Intergalactic Network to mean not just a group of people to whom he was sending memos but a universe of interconnected computers over which all might send their memos.

Licklider was no exception to the rule that people didn't spend a long time at ARPA. But by the time he left in 1964, he had succeeded in shifting the agency's emphasis in computing R&D from a command systems laboratory playing out war-game scenarios to advanced research in time-sharing systems, computer graphics, and improved computer languages. The name of the office, Command and Control Research, had changed to reflect that shift, becoming the Information Processing Techniques Office. Licklider chose his successor, a colleague named Ivan Sutherland, the world's leading expert in computer graphics. In 1965 Sutherland hired a young hotshot named Bob Taylor, who would soon sit down in ARPA's terminal room and wonder why, with so many computers, they were unable to communicate with one another.

Bob Taylor had started college in Dallas at the age of sixteen thinking he would follow his father's footsteps and become a minister. His family never lived in one place very long, moving from one Methodist church to another around Texas, in towns with names like Uvalde, Victoria, and Ozona. But instead of the service of the Lord, Taylor entered the service of the U.S. Navy when his reserve unit was called to active duty for the Korean War.

Taylor spent the war at the Dallas Naval Air Station—“the USS
Neverfloat,
” he called it. At the end of the war, he entered the University of Texas on the GI Bill with no particular course of study in mind. He finally graduated in 1957 with a major in psychology and a minor in mathematics.

Taylor pursued his love for science into graduate school at UT and wrote his dissertation on psychoacoustics, a field from which some people were making a leap into computing. Taylor jumped, too. “When I was in graduate school, there was no computer science,” Taylor recollected, “so I didn't really have much of an introduction to computing. But I began to feel that computing research was going to be much more rewarding.”

Fresh from school, Taylor held a few jobs in the aerospace industry before landing a job at NASA in 1961, where he worked as a program officer in Washington, D.C., in the Office of Advanced Research and Technology. One day in 1963 Taylor was invited to join an unofficial committee of government program managers, all of whom were involved in funding computer research. It was an informal group that simply exchanged information about their projects, looked for ways of collaborating, and tried to avoid duplication or overlap. The invitation had come, as it turned out, from someone who had been Taylor's intellectual role model in psychoacoustics—J. C. R. Licklider. He was head of the committee. Licklider's early work in psychoacoustics had deeply influenced Taylor's own and he welcomed the chance to meet the illustrious Licklider.

Taylor was struck by how unassuming he was. “He flattered me right off by saying he knew my thesis work,” Taylor said. Here was a man with a giant reputation who was probably one of the nicest, most easygoing people Taylor had ever met.

When Taylor first joined the committee, Licklider was in the process of pulling together the computer science community under ARPA, the new generation of researchers drawn to interactive computing. They were busily framing their bold new perspective, radically different from the mainstream in computer research and development during the previous two decades. Mountains of money and years of work had been invested in improving the technical parameters of speed, reliability, and memory size of computers. But this small avant-garde of researchers concentrated at MIT and around Boston had begun working on making the computer an amplifier of human potential, an extension of the mind and body.

Taylor was known for having a good bit of intuition himself. He was considered a farsighted program officer who had a knack for picking innovative winners—both projects and researchers. He joined ARPA in early 1965, following Licklider's departure, to work as deputy to Ivan Sutherland, IPTO's second director. Months later, in 1966, at the age of thirty-four, Taylor became the third director of IPTO, inheriting responsibility for the community and much of the vision—indeed the very office—established by Licklider. The only difference, which turned out to be crucial, was that ARPA—now headed by Charles Herzfeld, an Austrian physicist who had fled Europe during the war—was even faster and looser with its money than it had been during Ruina's tenure. A joke now circulated among its program directors: Come up with a good idea for a research program and it will take you about thirty minutes to get the funding.

The “terminal problem,” as Taylor called it, was a source of frustration not just for him but for Sutherland before him and for Licklider before that. One day, shortly after becoming IPTO director, Taylor found himself rolling around an idea Lick had discussed with him several times but never actually acted on. Now that it was his watch, Taylor decided to act.

Taylor headed straight to Herzfeld's office. No memos. No meetings. Other program directors were slightly intimidated by Herzfeld, a large man with a thick rumbling Viennese accent. But Taylor saw nothing to fear about the man. In fact, Taylor behaved like such a good old boy around his boss that someone once asked him,“Taylor, what have you got with Herzfeld? You must be related to Lyndon Johnson. You're both from Texas, aren't you?”

Taylor told the ARPA director he needed to discuss funding for a networking experiment he had in mind. Herzfeld had talked about networking with Taylor a bit already, so the idea wasn't new to him. He had also visited Taylor's office, where he witnessed the annoying exercise of logging on to three different computers. And a few years earlier he had even fallen under the spell of Licklider himself when he attended Lick's lectures on interactive computing.

Taylor gave his boss a quick briefing: IPTO contractors, most of whom were at research universities, were beginning to request more and more computer resources. Every principal investigator, it seemed, wanted his own computer. Not only was there an obvious duplication of effort across the research community, but it was getting damned expensive. Computers weren't small and they weren't cheap. Why not try tying them all together? By building a system of electronic links between machines, researchers doing similar work in different parts of the country could share resources and results more easily. Instead of spreading a half dozen expensive mainframes across the country devoted to supporting advanced graphics research, ARPA could concentrate resources in one or two places and build a way for everyone to get at them. One university might concentrate on one thing, another research center could be funded to concentrate on something else, but regardless of where you were physically located, you would have access to it all. He suggested that ARPA fund a small test network, starting with, say, four nodes and building up to a dozen or so.

The Defense Department was the largest buyer of computers in the world. Investing in a particular make of computer was no trivial decision, and it often put the different services in a bind, particularly when faced with a federal rule dictating that all manufacturers be given equal opportunity. There seemed to be no hope of curtailing the purchase of a whole variety of machines. And the chances seemed slim to nonexistent that the computing world would gravitate anytime soon to a set of uniform operating standards. Research sponsors like ARPA would just have to find some other way of overcoming the industry's incompatibility problems. If the network idea worked, Taylor told Herzfeld, it would be possible for computers from different manufacturers to connect, and the problem of choosing computers would be greatly diminished. Herzfeld was so taken with that possibility that those arguments alone might have been enough to convince him. But there was another advantage, centering on the question of reliability. It might be possible to connect computers in a network redundantly, so that if one line went down, a message could take another path.

“Is it going to be hard to do?” Herzfeld asked.

“Oh no. We already know how to do it,” Taylor responded with characteristic boldness.

“Great idea,” Herzfeld said. “Get it going. You've got a million dollars more in your budget right now. Go.”

Taylor left Herzfeld's office on the E-ring and headed back to the corridor that connected to the D-ring and his own office. He glanced at his watch. “Jesus Christ,” he said to himself softly. “That only took twenty minutes.”

A Block Here, Some Stones There

By the time Taylor assumed the directorship of IPTO in 1966, manifestations of Licklider's philosophy were evident throughout the computer research establishment. The ranks of researchers hoping to extend the computer beyond the status of a calculating instrument continued to grow throughout the decade. Some of the earliest and most important work in interactive graphics and virtual reality was taking place at the University of Utah using ARPA money. MIT in particular, seemed to breed one groundbreaking development after another. There Marvin Minsky and Seymour Papert were engaged in important early work in artificial intelligence. Programs at other institutions focused on advanced programming techniques, time-sharing, and computer languages.

Building a network as an end in itself wasn't Taylor's principal objective. He was trying to solve a problem he had seen grow worse with each round of funding. Researchers were duplicating, and isolating, costly computing resources. Not only were the scientists at each site engaging in more, and more diverse, computer research, but their demands for computer resources were growing faster than Taylor's budget. Every new project required setting up a new and costly computing operation. Depending on the computer being used and the number of graduate students being supported, IPTO's individual grants ranged from $500,000 to $3 million.

And none of the resources or results was easily shared. If the scientists doing graphics in Salt Lake City wanted to use the programs developed by the people at Lincoln Lab, they had to fly to Boston. Still more frustrating, if after a trip to Boston people in Utah wanted to start a similar project on their own machine, they would need to spend considerable time and money duplicating what they had just seen. In those days, software programs were one-of-a-kind, like original works of art, and not easily transferred from one machine to another. Taylor was convinced of the technical feasibility of sharing such resources over a computer network, though it had never been done.

Beyond cost-cutting, Taylor's idea revealed something very profound. A machine's ability to amplify human intellectual power was precisely what Licklider had had in mind while writing his paper on human-machine symbiosis six years earlier. Of course, Licklider's ideas about time-sharing were already bearing fruit at universities all over the country. But the networking idea marked a significant departure from time-sharing. In a resource-sharing network, many machines would serve many different users, and a researcher interested in using, say, a particular graphics program on a machine two thousand miles away would simply log on to that machine. The idea of one computer reaching out to tap resources inside another, as peers in a collaborative organization, represented the most advanced conception yet to emerge from Licklider's vision.

Taylor had the money, and he had Herzfeld's support, but needed a program manager who could oversee the design and construction of such a network, someone who not only knew Licklider's ideas but believed in them. This person had to be a first-rate computer scientist, comfortable with a wide range of technical issues.

How it was to be achieved didn't concern Taylor greatly, as long as the network was reliable and fast. Those were his priorities. Interactive computing meant you'd get a quick response from a computer, so in the modern computing environment it made sense that a network also should be highly responsive. And to be useful, it had to be working anytime you needed it. Whoever designed such a network needed to be an expert in telecommunications systems as well. It wasn't an easy combination to find. But Taylor already had someone in mind: a shy, deep-thinking young computer scientist from the Lincoln Labs breeding ground named Larry Roberts.

In early 1966, Roberts was at Lincoln working on graphics. But he had also done quite a lot of work in communications. He had just completed one of the most relevant proof-of-principle experiments in networking to date, hooking together two computers a continent apart. Taylor had funded Roberts's experiment. It had been successful enough to build Taylor's confidence and convince both himself and Herzfeld that a slightly more intricate network was feasible. And Roberts's knowledge of computers went deep. The son of Yale chemists, Roberts had attended MIT and received his introduction to computers on the TX-0. Although it was the first transistorized digital computer, the TX-0 was limited (subtraction was not in its repertoire; it could subtract only by adding a negative number). Using the TX-0, Roberts taught himself the basics of computer design and operation. Roberts, in fact, had written the entire operating system for its successor, the TX-2 computer at Lincoln, which Wes Clark (who built the TX-0 with Ken Olsen) had fatefully shown off to Licklider. When Clark left Lincoln in 1964, the job of overseeing the TX-2 had fallen to Roberts.

Taylor didn't know Roberts very well. No one, it seemed, knew Roberts very well. He was as reserved in his manner as Taylor was open in his. The people with whom Roberts worked most closely knew almost nothing about his personal life. What was known about him was that in addition to computing and telecommunications expertise, he had a knack for management. Roberts's style was simple, direct, unambiguous, and terribly effective.

Roberts had a reputation for being something of a genius. At twenty-eight, he had done more in the field of computing than many scientists were to achieve in a lifetime. Blessed with incredible stamina, he worked inordinately late hours. He was also a quick study: More than a few people had had the experience of explaining to Roberts something they had been working on intensively for years, and finding that within a few minutes he had grasped it, turned it around in his head a couple of times, and offered trenchant comments of his own. Roberts reminded Taylor of Licklider a little—but without Lick's sense of humor.

Roberts was also known for his nearly obsessive ability to immerse himself in a challenge, pouring intense powers of concentration into a problem. A colleague once recalled the time Roberts took a speed-reading course. He quickly doubled his already rapid reading rate, but he didn't stop there. He delved into the professional literature of speed-reading and kept pushing himself until he was reading at the phenomenal rate of about thirty thousand words a minute with 10 percent “selective comprehension,” as Roberts described it. After a few months, Roberts's limiting factor had nothing to do with his eyes or his brain but with the speed at which he could turn the pages. ”He'd pick up a paperback and be through with it in ten minutes,” the friend observed. “It was typical Larry.”

Taylor called Roberts and told him he'd like to come to Boston to see him. A few days later Taylor was sitting in Roberts's office at Lincoln Lab, telling him about the experiment he had in mind. As Taylor talked, Roberts murmured a nasal “hmm-hmm” as if to say, “please go on.” Taylor outlined not just the project but a job offer. Roberts would be hired as program director for the experimental network, with the understanding that he would be next in line for the IPTO directorship. Taylor made it clear that this project had the full support of ARPA's director and that Roberts would be given ample latitude to design and build the network however he saw fit. Taylor waited for an answer. “I'll think about it,” Roberts said flatly.

Taylor read this as Roberts's polite way of saying no, and he left Boston discouraged. Under any other circumstances, he'd have simply crossed Roberts off the list and called his second choice. But he didn't have a second choice. Not only did Roberts have the necessary technical understanding, but Taylor knew he would listen to Licklider and Wes Clark, both of whom were supporting Taylor's idea.

A few weeks later Taylor made a second trip to Lincoln. This time Roberts was more forthcoming. He told Taylor politely but unequivocally that he was enjoying his work at Lincoln and had no desire to become a Washington bureaucrat.

Disconsolate, Taylor went to Cambridge to visit Lick, who was now back at MIT ensconced in a research effort on time-sharing called Project MAC. They discussed who else might be well suited to the job. Lick suggested a few people, but Taylor rejected them. He wanted Roberts. From then on, every two months or so, during visits to ARPA's other Boston-area contractors, Taylor called on Roberts to try to persuade him to change his mind.

It had been nearly a year since Taylor's twenty-minute conversation with Herzfeld, and the networking idea was floundering for lack of a program manager. One day in late 1966, Taylor returned to the ARPA director's office.

“Isn't it true that ARPA is giving Lincoln at least fifty-one percent of its funding?” Taylor asked his boss.

“Yes, it is,” Herzfeld responded, slightly puzzled.

Taylor then explained the difficulty he was having getting the engineer he wanted to run the networking program.

“Who is it?” Herzfeld asked.

Taylor told him. Then he asked his boss another question. Would Herzfeld call the director of Lincoln Lab and ask him to call Roberts in and tell him that it would be in his own best interest—and in Lincoln's best interest—to agree to take the Washington job?

Herzfeld picked up his telephone and dialed Lincoln Lab. He got the director on the line and said just what Taylor had asked him to say. It was a short conversation but, from what Taylor could tell, Herzfeld encountered no resistance. Herzfeld hung up, smiled at Taylor, and said, “Well, okay. We'll see what happens.” Two weeks later, Roberts accepted the job.

Larry Roberts was twenty-nine years old when he walked into the Pentagon as ARPA's newest draftee. He fit in quickly, and his dislike of idle time soon became legendary. Within a few weeks, he had the place—one of the world's largest, most labyrinthine buildings—memorized. Getting around the building was complicated by the fact that certain hallways were blocked off as classified areas. Roberts obtained a stopwatch and began timing various routes to his frequent destinations. ”Larry's Route” soon became commonly known as the fastest distance between any two Pentagon points.

Even before his first day at ARPA, Roberts had a rudimentary outline of the computer network figured out. Then, and for years afterward as the project grew, Roberts drew meticulous network diagrams, sketching out where the data lines should go, and the number of hops between nodes. On tracing paper and quadrille pad, he created hundreds of conceptual and logical sketches like these: