Where Wizards Stay Up Late: The Origins of the Internet

by Katie Hafner

  They found that Secretary McElroy had a similar appreciation for them. One aspect of his career at P&G that he was most proud of was the amount of money the company had devoted to research. He believed in the value of unfettered science, in its ability to produce remarkable, if not always predictable, results. McElroy and P&G had created a large “blue-sky” research laboratory, had funded it well, and rarely if ever pressured its scientists to justify their work. It was one of the first corporate research operations of its kind, one in which scientists were left to pursue almost anything and were well supported from the top.

  Significant technological advances had come from a similar arrangement between universities and government during World War II: radar, nuclear weapons, and large calculating machines resulted from what Killian called “the freewheeling methods of outstanding academic scientists and engineers who had always been free of any inhibiting regimentation and organization.”

  In consultation with Killian, whose support was crucial, McElroy began discussing the idea of establishing an independent agency for research. Perhaps McElroy was aware that the U.S. Chamber of Commerce had floated the notion of creating a single research-anddevelopment agency for the federal government during congressional hearings months beforeSputnik.Such talk was in the air. The idea now emerged in discussions with an informal advisory committee of astute industrialists who met regularly with the secretary of defense.

  In the days immediately following the Soviet launch, two men had been to see McElroy: eminent nuclear physicist Ernest O. Lawrence, and Charles Thomas, a former CEO of the Monsanto Chemical Company and an occasional advisor to the president. In a meeting lasting several hours, they discussed the idea of a strong advanced R&D agency that would report to the secretary, and both visitors urged McElroy to run with it. Physicist Herbert York, director of the Livermore Laboratory and close confidant to both Eisenhower and Killian, joined the conversations. At the same time, McElroy himself was consulting frequently with Killian and the president. In Killian’s view, the traditional missions of the armed services had been outmoded by modern science and technology. Here was a way to move the Pentagon into the new age. One of the principal attractions of the research agency concept, in McElroy’s mind, was the ability it would give him to manage the fierce competition within DOD over R&D programs and budgets. The competition was reaching absurd new heights. Army, Navy and Air Force commanders treatedSputniklike the starting gun in a new race, each vying against the other for the biggest share of R&D spending.

  McElroy believed that a centralized agency for advanced research projects would reduce interservice rivalry by placing federal R&D budgets substantially under his own close supervision. Moreover, it would almost certainly appeal to the president, since the military’s often selfish claims and special-interest hype usually met with disdain in the White House.

  Eisenhower liked the secretary’s proposal. While the administration had larger plans in store—the creation of the National Aeronautics and Space Administration (NASA), passage of a Defense Reorganization Act, establishment of the office of a Director of Defense Research and Engineering—those things would take time. The Advanced Research Projects Agency concept was a response the president could point to immediately. And it would give him an agile R&D agency he and his scientists could call on in the future.

  On November 20, 1957, McElroy went to Capitol Hill for the first time as secretary. In the course of testifying about U.S. ballistic missile programs he mentioned that he had decided to set up a new “single manager” for all defense research. In the near term, he told Congress, the new agency would handle satellite and space R&D programs. The next day he circulated to the Joint Chiefs of Staff a draft charter and directive for the new agency, asking for their review and comments.

  The military bristled at the implied criticism. It was one thing to have civilian science and technologyadvisors,but McElroy was now clearly invading their turf, planning to manage and operate a central office that would control defense R&D programs and future weapons projects. They were prepared to fight back. Secretary of the Air Force James Douglas wrote, “The Air Force appreciates that the proposals are suggestions.” That is, message received, not accepted. The Army, Navy, and Joint Chiefs of Staff returned the draft charter with numerous revisions. They fiddled deviously with language concerning the new agency’s contracting authority, their revisions peppered with subversive word changes, crafty additions, and sly deletions all designed to weaken and confine the agency.

  McElroy conceded a few minor points but made two things clear: The agency’s director would have contracting authority, and the scope of the agency’s research was to be unlimited.

  On January 7, 1958, Eisenhower sent a message to Congress requesting startup funds for the establishment of the Advanced Research Projects Agency. Two days later he drove the point home in his State of the Union Message. “I am not attempting today to pass judgment on the charges of harmful service rivalries. But one thing is sure. Whatever they are, America wants them stopped.”

  Eisenhower also affirmed “the need for single control in some of our most advanced development projects,” then delivered his coup de grâce to the generals: “Another requirement of military organization is a clear subordination of the military services to the duly constituted civilian authority. This control must be real; not merely on the surface.”

  In early 1958, Roy Johnson, a vice president of General Electric, was appointed ARPA’s first director; five days after the appointment, funding was approved by Congress (as a line item in an Air Force appropriations bill), and ARPA went into business.

  In the postSputnikpanic, the race for outer space cut a wide swath through American life, causing a new emphasis on science in the schools, worsening relations between Russia and the United States, and opening a floodgate for R&D spending. Washington’s “external challenge” R&D spending rose from $5 billion per year to over $13 billion annually between 1959 and 1964.Sputniklaunched a golden era for military science and technology. (By the mid-1960s, the nation’s total R&D expenditures would account for 3 percent of the gross national product, a benchmark that was both a symbol of progress and a goal for other countries.)

  All eyes were on ARPA when it opened its doors with a $520 million appropriation and a $2 billion budget plan. It was given direction over all U.S. space programs and all advanced strategic missile research. By 1959, a headquarters staff of about seventy people had been hired, a number that would remain fairly constant for years. These were mainly scientific project managers who analyzed and evaluated R&D proposals, supervising the work of hundreds of contractors.

  Roy Johnson, ARPA’s first director, was, like his boss, a businessman. At age fifty-two, he had been personally recruited by McElroy, who convinced him to leave a $160,000 job with General Electric and take an $18,000 job in Washington.

  Not surprisingly, Johnson approached America’s R&D agenda as a management problem. Management skills were his strong suit. His job, as he saw it, was to exhort people to do anything necessary to get the edge on the Soviets. He argued often and vigorously with rooms full of generals and admirals, and aggressively took on the Air Force. It soon became apparent that Johnson was a serious and vociferous advocate of a strong military presence in outer space.

  But of course Killian and other scientists around Eisenhower had wanted someone well versed in scientific and technological issues running ARPA. Johnson had been instructed to hire a military flag officer as his deputy, and to select a chief scientist to round out his leadership team. The scientific post had almost gone to Wernher von Braun, until he insisted on bringing his whole team of a dozen associates with him to the Pentagon. So Herbert York, whom Killian had been keen on, was given the job and moved to ARPA from the Lawrence Livermore Laboratory. When he arrived on the third floor at the Pentagon, York promptly hung a large picture of the moon on his office wall. And right beside it he hung an empty frame. He told visitors it would soon be filled with t
he first picture of the back side of the moon.

  The rest of ARPA’s staff was recruited from industry’s top-flight technical talent at places like Lockheed, Union Carbide, Convair, and other Pentagon contractors. The staff’s days were spent sifting for gold in a torrential flow of unsolicited R&D proposals.

  ARPA’s success hinged on Johnson’s extremely vocal posturing about America’s role in outer space and his simplistic view of Soviet-American tensions. He mistakenly defined ARPA’s mission almost entirely in military terms, outlining the kind of space projects he envisioned: global surveillance satellites, space defense interceptor vehicles, strategic orbital weapon systems, stationary communications satellites, manned space stations, and a moon base.

  Eisenhower and his civilian scientists had moved ahead with the rest of their agenda, and by the late summer of 1958 the National Aeronautics and Space Administration had been enacted into law. Almost overnight, while Johnson drummed for a military presence in space, the space projects and missile programs were stripped away from ARPA and transferred over to NASA or back to the services, leaving ARPA’s budget whittled to a measly $150 million. ARPA’s portfolio was gutted, its staff left practically without any role.Aviation Weekcalled the young agency “a dead cat hanging in the fruit closet.”

  Johnson resigned. Before leaving, however, he commissioned his staff to draft a paper weighing four alternatives: abolishing ARPA, expanding it, making no changes, or redefining its mission. His analysts threw themselves into building a case for the fourth alternative, and their tenacity alone saved the agency from sure oblivion. They fleshed out a set of goals to distance ARPA from the Pentagon by shifting the agency’s focus to the nation’s long-term “basic research” efforts. The services had never been interested in abstract research projects; their world was driven by near-term goals and cozy relationships with industrial contractors. The staff of ARPA saw an opportunity to redefine the agency as a group that would take on the really advanced “far-out” research.

  Most important of all, ARPA staffers recognized the agency’s biggest mistake yet: It had not been tapping the universities where much of the best scientific work was being done. The scientific community, predictably, rallied to the call for a reinvention of ARPA as a “high-risk, high-gain” research sponsor—the kind of R&D shop they had dreamed of all along. Their dream was realized; ARPA was given its new mission.

  As ARPA’s features took shape, one readily apparent characteristic of the agency was that its relatively small size allowed the personality of its director to permeate the organization. In time, the “ARPA style”—freewheeling, open to high risk, agile—would be vaunted. Other Washington bureaucrats came to envy ARPA’smodus operandi. Eventually the agency attracted an elite corps of hard-charging R&D advocates from the finest universities and research laboratories, who set about building a community of the best technical and scientific minds in American research.

  The agency’s new basic research and special-project orientation was ideally suited to the atmospheric change in Washington caused by the election of John F. Kennedy. With vigor, Washington’s bureaucracies responded to the Kennedy charisma. At the Pentagon, Robert S. McNamara, the new secretary of defense, led the shift away from the philosophy of “massive retaliation” in America’s strategic posture, and toward a strategy of “flexible response” to international threats to American supremacy. Science was the New Frontier.

  In early 1961 ARPA’s second director, Brigadier General Austin W. Betts, resigned and was replaced by Jack P. Ruina, the first scientist to direct ARPA. Ruina came with strong academic credentials as well as some military background. Trained as an electrical engineer, he had been a university professor and had also served as a deputy assistant secretary of the Air Force. He was on good terms with members of the science advisory panels to the White House.

  A golden era for ARPA was just beginning. Ruina brought a relaxed management style and decentralized structure to the agency. Details didn’t interest him; finding great talent did. He believed in picking the best people and letting them pick the best technology. He felt strongly about giving his program directors free rein. His job, as he saw it, was to build as much support and funding as he could for whatever projects they selected. Ruina had a theory that truly talented people wouldn’t normally choose to hang around long in a government bureaucracy but could be convinced to spend a year or two if offered enough flexibility and large enough budgets. Turnover didn’t bother him. The agency, he believed, would benefit from frequent exposure to fresh views. In keeping with all that, he saw himself as a short-timer, too.

  In time, Ruina raised ARPA’s annual budget to $250 million. Projects in ballistic missile defense and nuclear test detection, couched in terms of basic research, were the top priorities. (There were also programs like behavioral research and command and control which, though interesting, fell below the level of Ruina’s close attention.)

  Then in May 1961 a computer, and a very large one at that, demanded his attention. A program in military command and control issues had been started at ARPA using emergency DOD funds. For the work, the Air Force had purchased the huge, expensive Q-32, a behemoth of a machine that was to act as the backup for the nation’s air defense early-warning system. The machine had been installed at a facility in Santa Monica, California, at one of the Air Force’s major contractors, System Development Corporation (SDC), where it was supposed to be used for operator training, and as a software development tool.

  Then the Air Force had been forced to cut back, which left the Santa Monica firm begging for some way to keep its employees working with the computer. The Air Force, which had already sunk millions into the SDC contract, suddenly had an embarrassing white elephant on its hands in the form of one massive, ungainly computing machine.

  Licklider

  The relationship between the military and computer establishments began with the modern computer industry itself. During World War II, in the midst of a perceived need for faster calculating ability than could be provided by the banks of mechanical calculators run by human operators, the military funded dozens of computing experiments. The Navy supported Howard Aiken, the Harvard mathematics professor who dreamed of building a large-scale calculator and ended up with the Mark I, a fiftyone-foot-long, eight-foot-tall switchboard that could perform arithmetical operations without the intervention of an operator. The Army also supported the famous Electronic Numerical Integrator And Calculator (ENIAC) project at the University of Pennsylvania. Later at MIT, first the Navy and then the Air Force supported a computer called Whirlwind.

  In the early 1950s computing meant doing arithmetic fast. Companies, especially banks, put their machines to work doing large-scale calculations. In 1953, International Business Machines Corporation (IBM), already the country’s largest manufacturer of time clocks as well as electromechnical tabulating equipment, jumped into the business of making large electronic computers. These were business machines of the future. The IBM machines weren’t necessarily better than the Univac (the successor to the ENIAC), but IBM’s sales staff became legendary, and before too long sales of IBM’s machines had surpassed those of the Univac.

  Then, in the late 1950s, just as IBM was passing the billion-dollar-sales mark, Ken Olsen, an individualistic and outspoken engineer, left MIT’s Lincoln Laboratory with $70,000 in venture capital to exploit commercially the technology developed around a new machine: the TX-2 at Lincoln Lab. He formed the Digital Equipment Corporation to manufacture and sell computer components, and then he built something radically different from what had existed before: a smaller computer called a minicomputer that interacted directly with the user. Olsen’s idea for an interactive computer had come from a pioneering group of computer researchers at MIT. A different, slightly younger group there came up with another dramatic concept in computing that was beginning to catch on, particularly in academic institutions. They called it “time-sharing,” and it had obvious appeal as an alternative to the slow a
nd awkward traditional method of “batch” processing.

  Batch processing was a cumbersome way of computing. For even the smallest programming task, it was necessary to have the relevant code punched onto program cards, which were then combined with “control cards” to take care of the computer’s administrative functions. A computer operator fed the cards into the computer or onto magnetic tape for processing, one batch at a time. Depending on the length of the queue and the complexity of the programs and problems, the wait could be long. It was not unusual to wait a day or longer for results.

  Time-sharing was, as the term suggests, a new method of giving many users interactive access to computers from individual terminals. The terminals allowed them to interact directly with the mainframe computer. The revolutionary aspect of time-sharing was that it eliminated much of the tedious waiting that characterized batch-process computing. Time-sharing gave users terminals that allowed them to interact directly with the computer and obtain their results immediately. “We really believed that it was a better way to operate,” recalled Fernando Corbató, an MIT computer scientist. “I suppose if somebody had said, ‘I will give you a free machine,’ we might have said, ‘We do not need time-sharing.’” But computers in those days were huge things. They took up large rooms and required continual maintenance because there were so many components. “They were just not a casual thing,” Corbató went on. “You did not normally think of having a personal machine in those days—exclusive use maybe, but not a personal one. So we really saw a need to try to change that. We were frustrated.” The appreciation of time-sharing was directly proportional to the amount of direct access one had to the computer. And usually that meant that the more you programmed, the better you understood the value of direct access.