(Corrects Baran's date of death.)
In 1960, the Rand Corp. published a 24-page paper, opaquely titled "Reliable Digital Communications Systems Using Unreliable Network Repeater Nodes." The paper came as a response to a problem that was vexing the U.S. Air Force: How could the President send orders to the commanders of his missile silos after a first strike by the Soviets? Thirty years later, in an oral history with the Charles Babbage Institute at the University of Minnesota, the paper's author, Paul Baran, explained that his intention was to prevent the U.S. from launching a preemptive attack. He saw the arms race as a dangerous test of game theory: A country worried about the survival of its command and control system was a country with a jittery trigger finger. Baran, who died on Mar. 26, imagined the Internet in order to save the world from nuclear annihilation.
He would demur, of course; he liked to say of technological progress that Peter and Paul each laid stones that became a cathedral. But his metaphor suggests a less modest truth, that he stood among the holy few government-funded academics who created the Internet. And the lessons of his academic work—though with his master's in engineering, he protested that his wife, not he, was the only Dr. Baran—still inform how we think about and manage the Internet today.
Baran opened his Rand paper by considering the day after the bombs fell. "Plan to minimize potential destruction and to do all those things necessary," he wrote, "to permit the survivors of the holocaust to shuck their ashes and reconstruct the economy swiftly." We now have think tanks and consulting firms; in the beginning there was Rand—and only Rand could have produced that line. World War II had proved the value of bringing scientists together to solve problems, and in 1946, Douglas Aircraft put a group of thinkers in a building in downtown Santa Monica, Calif., under a single yearly contract from the Air Force to think the unthinkable. In Paul Baran's day, the people of Rand Corp. loomed, brainy and calculating, in American movies and song.
They wore white shirts with skinny ties and used a slide rule called the "Rand bomb damage effect calculator" to estimate megadeaths. They also sat in beanbag chairs and rattan lawn furniture. They built a tunnel underneath the Pacific Coast Highway to the beach. The futurist Herman Kahn, one of the models for the character of Dr. Strangelove, worked there. So did Daniel Ellsberg, who in a crisis of conscience in the early 1970s leaked much of his Vietnam War work at Rand—a trove that became known as the Pentagon Papers. Rand's one-client business gave it a purpose, but the nature of its contract with the Air Force also gave it an extraordinary amount of freedom to pursue strange ideas. James Thomson, Rand's current president, says people used to say at the company that the Air Force "threw money over the transom to see what came out the door." Rand employees chose what they wanted to work on; Baran picked the survival of networks.
He started by abandoning the idea of perfection. AT&T (T), then the country's telephone monopoly, designed its network to make the most efficient use of every element. Yet the system's centralized structure left it vulnerable to attack. Baran assumed that, in the event of a nuclear war, any single point—or "node"—in a communications network carried a high probability of failure. He proposed extra capacity, and an architecture that "makes itself content with the assured operation of a certain fraction of the network."
Then he hied to the living room couch. He posited that if he absolutely had to catch the cowboys-and-cardsharps television show Maverick, he might buy a second TV. If each TV had a 0.001 percent chance of failure, there was a 0.000001 percent chance that he would miss his show. To make his theory more relevant, he asked the reader to imagine his network as members of Congress voting from their home offices. He would measure success by imagining an attack and comparing the number of surviving congressmen to the number of survivors able to talk to each other. Then, as they also learned to do at Rand, he ran scenarios.
Arrange your congressmen in equal columns and connect them all with a single, snaking line. This is the most efficient network structure. There is roughly one link for every congressman, or node. Baran defines the ratio as "redundancy"; this network has a redundancy approaching one. Next, connect all nodes vertically and horizontally, creating a grid, and your network has a redundancy of two. Now, add diagonal lines in both directions through your grid, and you've reached four. Baran ran simulations in which he dispatched varying numbers of congressmen. He drew a curve showing that a redundancy of three or four offered results close to a theoretical limit of politicians who'd still be alive. This meant that slightly greater redundancy offered much greater stability. Already, in this first paper, his higher-redundancy networks resemble what we now know as the Internet.
Baran showed his work around and began answering questions. His network presupposed several technological advances we now take for granted. Computers were large and expensive, and AT&T's efficient networks put the switching—the intelligence—at the center. A redundant network would have to decide how to route around damage, which meant that every node would have to be smarter. Paul Saffo, a futurist and a friend of Baran's, points out that in the paper, "Paul's talking about clock speed. He's anticipating what we know as Moore's Law," the now common understanding that computing power doubles every two years. Baran suggested cheaper, smaller computers to make decisions at every node—these devices would come to be called "routers."
Talking to AT&T—which would have to carry his network—Baran grew used to the condescension of communications engineers who responded with what he called "the melodic refrain of 'bulls--t.' " Tom Wheeler, a future friend and business partner, says Baran eventually learned to ask computer scientists from Bell Labs to serve as interpreters. "He would say something," says Wheeler, "and the Bell Labs guy would translate it into a Western Electric parts number, and the AT&T guys would go, 'Ooooh.' "
Baran also discovered, while killing imaginary members of Congress, that the most vulnerable nodes—the ones with the fewest links—lay at the edges of his network. We see this confirmed today as governments filter and block online access; the fewer connections that run into a country, the easier it is to cut it off from the rest of the Internet. It's no accident that all of China's almost 200 million IP addresses reach the Internet through only four nodes.
He turned the answers to the questions he encountered into a 13-volume set, detailing the arguments for a technique others would call "packet switching": discrete bundles of information stumbling between nodes, a departure from AT&T's rigid telephone circuits. Rand Corp. did not take out a patent on Baran's work, and left 11 volumes unclassified. "The focus of all those that I knew were concerned about the nation's defense was on avoidance of war," he said in his Charles Babbage Institute interview. "I never encountered anyone who deserved the Dr. Strangelove war-monger image." The U.S. would be safer if it knew its command structure would survive. It would be even safer, he argued, if the Soviet Union read his work and had the same network.
He never built it. After a reorganization at the Pentagon, Baran feared that the group responsible for communications wouldn't be able to pull it off, so in 1966 he recommended that his system not be built. "If the project turned into a botch," he said, "it would be extremely difficult to get it going again." That would wait until the Advanced Research Project Agency's network, which went live almost a decade after his first paper.
Accounts vary on how much the ARPAnet relied on Baran's work. It's safe to say, at least, that he envisioned something with no precedent. "There was a void," says Wheeler, "a total knowledge vacuum, and somehow he put something into it. That is what is just amazing."
Baran left Rand in 1968. He founded the Institute for the Future and seven companies; five went public. He worked to apply packet switching to cable television, satellite signals, and radio. He developed technology that DSL relies on. And, sitting in bed before he died, Paul Baran sketched out an idea and sent it to Wheeler for a look. It was a detailed plan on how to run a better hospital.