A funding agency with a long view
The internet’s ancestor was not built by a telephone company or a computer maker. It was funded by a United States government research agency, the Advanced Research Projects Agency, known as ARPA (later DARPA). ARPA was created in 1958 in the aftermath of Sputnik, and in the early 1960s its Information Processing Techniques Office began funding computing research at universities and other labs.
The first director of that office, J. C. R. Licklider, a psychologist and computer scientist, promoted the idea of interactive computing and, in an early-1960s memo, described a vision of connected machines that he playfully called an “Intergalactic Computer Network.” He left the office after a short period, but the idea passed on to successors. Ivan Sutherland, Robert Taylor and later Lawrence Roberts continued the effort. Taylor, who led the office in the mid-1960s, is often credited with spotting a practical problem: his office funded researchers at different sites who each had their own terminals and incompatible computers, and he wanted them to share resources without a pile of separate terminals in his own office. Roberts then took on the job of designing and managing the network.
It is worth noting a common misunderstanding. The network was not created to survive a nuclear war. That story is widely repeated, but the people involved have mostly said the ARPANET was meant to share computing resources among researchers. Separate earlier work on survivable communications, discussed in the history of packet switching, influenced the broader field and is the likely source of the confusion.
Building the first links
ARPA asked for bids to build the small computers that would sit at each site, and the contract went to the Cambridge, Massachusetts firm Bolt Beranek and Newman, known as BBN. Its team, including Frank Heart, Robert Kahn and others, built the Interface Message Processors, using ruggedized minicomputers. Host sites had to write software to connect their own machines to the interface, and researchers organized a Network Working Group, with younger scholars such as Steve Crocker taking leading roles, to agree on host-to-host protocols.
The first four nodes were:
- the University of California, Los Angeles, where Leonard Kleinrock’s group had a measurement role
- the Stanford Research Institute, where Douglas Engelbart’s team ran an online system for collaborative work
- the University of California, Santa Barbara
- the University of Utah
The first connection between UCLA and SRI is usually dated to late October 1969, and the four-node network was working by the end of that year. A widely repeated detail is that the first attempted login message failed partway through, though accounts vary on exactly how far it got.
Growth through the 1970s
New sites joined during the early 1970s: other universities, government laboratories, and companies working on contract. A public demonstration at a computer communications conference in 1972 helped show the technology to a wider audience. Around then, the network gained the application that most changed how people used it. Ray Tomlinson, an engineer at BBN, is credited with adapting an electronic mail program to send messages between machines on the network in 1971 and with choosing the @ sign to separate user and host names. Email quickly became the network’s most popular use, and researchers began using mailing lists to hold discussions. The longer story is told in the history of electronic mail.
Meanwhile other communities built parallel systems. Usenet, a discussion network that spread over telephone connections among Unix machines, began around 1979 and 1980 and reached people who were not on the ARPANET. In the early 1980s, a network called CSNET extended connections to computer science departments that did not have ARPA funding.
The shift to internetworking
During the 1970s, the leaders of the effort, including Vint Cerf and Robert Kahn, concluded that the ARPANET should connect to other kinds of networks, such as radio and satellite networks. They published a design in 1974 that grew into TCP/IP. After a period of testing, the ARPANET changed over to the new protocols at the start of 1983. In the following years, the National Science Foundation funded a national academic backbone, and gradually the ARPANET itself was retired from use around 1990. For the technical side of this transition, see the article on the early days of computer networking, along with the one on internet protocols.
Whose achievement?
It would be tidy to credit one inventor, but the early internet was assembled by many hands: program managers who found money, engineers who built the hardware, graduate students who wrote and documented protocols, and committees that argued about details. Rival network projects in Britain and France offered ideas that fed back into American work.
A culture also emerged. Technical documents were circulated as informal Requests for Comments, which anyone could write and critique. Working code mattered more than formal credentials. That open, collegial style helped the network grow, and it also reflected a small group that shared assumptions about trust. As the network grew, those assumptions would be tested by problems of security and abuse that the founders had not planned for.