Forty years ago today, a team led by Leonard Kleinrock typed the “Lo” of “Login” into a Stanford computer, which promptly crashed before the command could be entered. But because Kleinrock’s team was sending this message from a UCLA machine, he had just taken part in one of the great milestones in communication history.*
These computers were connected under the auspices of the Advanced Research Projects Agency (ARPA). The agency was created by President Eisenhower as part of the Department of Defense in 1958 as a direct response the Soviet Union’s launch of the first satellite, Sputnik, in 1957. Its mission: to ensure that the United States never again be caught off guard by technological advances. Now known as the Defense Advanced Research Projects Agency (DARPA), the agency continues to push the envelope in such diverse fields as advanced propulsion, medicine, and robotics, as well as information technology. But it was ARPA—specifically its Information Processing Techniques Office (IPTO)—that funded the 1960s research project called ARPANET that became the internet of today.
In that era, even timesharing computer systems were few and far between, with only highlevel researchers interested in using them. These were unique machines or at least in very limited production runs, manufactured for specific purposes with software written expressly for them. Each machine had its cadre of technicians and programmers who communed with it in a process that seemed to draw more from the black arts of the occult than the realm of engineering.
There were no software manuals, no standardized operating systems. Each machine had its own input/output system, and there certainly was no standardized means of transmitting information from one machine to another. To create ARPANET, IPTO director Bob Taylor and the people he hired would have to create all of those things. They would have to, in fact, lay the foundations for what we know as the modern computer industry today.
In 1966, Taylor could readily see the need for a more comprehensive computer system. Each computer that Taylor could interact with in his office spoke its own language, required a different log-in procedure on a different terminal, and, of course, couldn’t talk to the others.
What he needed, thought Taylor, was a way to use just one terminal and one log-in for all three of the computers to which he had access. In other words, he needed a way for them to talk to one another. A computer network. And not just one that linked computers of the same type, as had already been done experimentally, but of different designs. Once he had that, the computer scientists around the country working on ARPA-funded projects could have it, too. They’d then be able to share computer resources and double, triple, even quadruple their computing power. With a computer network through which any computer could communicate with any other computer on the network, research could make much more efficient use of the then-still-very-expensive computer resources—they could all get more time on more computers.
Taylor described his brainstorm later as “kind of an ‘aha’ idea.” Right then and there, he headed over to the Pentagon’s E-Ring, to the office of Charlie Herzfeld, ARPA’s then-director, and pitched the idea to him. Herzfeld listened, asked a few questions, and then, as Taylor described it later, “He pretty much instantly made budget changes within his agency and took a million dollars away from one of his other offices and gave it to me to get started.” Back at his own office in the D-Ring, Taylor looked at his watch and let out a breath. “Jesus Christ,” he said to himself. “That took only 20 minutes.”
Watch Leonard Kleinrock describe the birth of the internet—and explain the meaning of the first two letters sent—in this video from UCLA.
Step one for Taylor was to hire a program manager for ARPANET. Larry Roberts, a twenty-something computer expert specializing in the field of computer graphics at MIT’s Lincoln Laboratory, home of Project SAGE, turned out to be the perfect fit for the job. Taylor had already funded a Robertsrun networking experiment during which Roberts and his team had successfully linked two similar computers thousands of miles apart. That made Roberts one of the few people in the world actually qualified to advance the field of computer networking.
The only problem was Roberts liked his job at MIT too much to leave it to become what he considered a mere Washington bureaucrat. Taylor spent an entire year working on Roberts, all the while not finding a suitable substitute. Finally, in desperation, he looked into the Lincoln Lab’s sources of funding, found that it got more than half from ARPA, and thereby got the leverage he needed to pry Roberts loose. At Taylor’s request, ARPA Director Herzfeld called Roberts’s boss at Lincoln Labs and explained how it would be in the lab’s best interest to let Roberts come work with him. That did the trick, and though Roberts went to ARPA grudgingly, he quickly immersed himself in his new work.
To solve the problem of getting disparate computer systems to talk to one another, ARPA researchers hit upon the solution of building the ARPANET out of identical gobetween computers whose only job would be to interpret data from the host computers and pass the data, properly translated, on to the network. Each host computer would be coupled to its own refrigerator-size interface message processor, or IMP.
The IMPs were also known as packet switches because of the scheme they used to pass on data. To protect against data loss associated with transmitting data over long distances, and to maximize the limited bandwidth available to them over the 1960s telephone network, the ARPANET’s engineers borrowed a technique proposed first by Leonard Kleinrock in his MIT PhD dissertation, and then independently by RAND Corporation engineer Paul Baran and Donald Davies in Britain. Baran arrived at the scheme with a way for the US communications network to survive a nuclear attack. Instead of taking a direct line of communication from a message’s source through a centralized distribution point to its destination, as the US phone network was designed to do, he proposed chopping each message into discrete chunks he called message blocks, routing each chunk through the communications network via different pathways, and reassembling them at their destination. “I get credit for a lot of things I didn’t do,” Baran said in a 2008 interview. “I just did a little piece on packet switching and I get blamed for the whole goddamned internet, you know?”
UCLA received the first IMP in September 1969. There, a team of faculty members and students attached it to their mainframe computer. Stanford Research Institute got an IMP of its own the following month, and on October 29, 1969, the first two nodes on the ARPANET (and the internet it birthed) communicated with each other up and down the California coast over 400 miles of leased telephone lines. In a glitch presaging the annoyance of internet users everywhere when experiencing dropped connections, the SRI machine accepted exactly two letters from the UCLA computer in that first connection before crashing.
As Leonard Kleinrock, the faculty member in charge of the project at UCLA, was later to note, that one small step for a computer and one giant leap for computerkind went completely unnoticed by the world at large. He said later, “1969 was quite a year. Man on the Moon. Woodstock. Mets won the World Series. Charles Manson starts killing these people here in Los Angeles. And the internet was born. Well, the first four everybody knew about. Nobody knew about the internet.” It would be decades before the significance of the achievement was widely recognized, and even today, the full potential of the globespanning network and the millions of nodes that it spawned is only just being realized.
Michael Belfiore is the author of The Department of Mad Scientists: How DARPA Is Remaking Our World, from the Internet to Artificial Limbs (available now), from which this article is adapted. Find Belfiore online and sign up for updates at www.michaelbelfiore.com.
*A previous version of this story implied that Kleinrock himself typed in the message, when it was actually another member of his research team, Charlie Kline.
Originally published October 29, 2009