ARPANET arose from a desire to share information over great distances without the need for dedicated phone connections between each computer on a network. As it turned out, fulfilling this desire would require “packet switching.”

Paul Baran, a researcher at the RAND Corporation think tank, first introduced the idea. Baran was instructed to come up with a plan for a computer communications network that could survive nuclear attack and continue functioning. He came up with a process that he called “hot-potato routing,” which later became known as packet switching.

“Packets” are small clusters of digital information broken up from larger messages for expediency’s sake. To illustrate in more recent terms: an e-mail might be split into numerous electronic packets of information and transmitted almost at random across the labyrinth of the nation’s telephone lines. They do not all follow the same route and do not even need to travel in proper sequential order. They are precisely reassembled by a modem at the receiver’s end, because each packet contains an identifying “header,” revealing which part of the larger message it represents, along with instructions for reconstituting the intended message. As a further safeguard, packets contain mathematical verification schemes that insure data does not get lost in transit. The network on which they travel, meanwhile, consists of computerized switches that automatically forward packets on to their destination. Data packets make computer communications more workable within existing telephone infrastructure by allowing all those packets to flow following paths of least resistance, thereby preventing logjams of digital data over direct, dedicated telephone lines.

Baran’s idea was ignored by the military. A 1964 paper outlining his innovation was published, but it was classified and began to collect dust. Fortunately, one place it was collecting dust was in the offices of ARPA, where it was eventually rediscovered. Baran’s idea became the key concept that made ARPANET possible. Packet-switched communication remains perhaps the most important legacy handed down to the Internet by ARPANET.

Rise and fall

In late 1969, a team of UCLA graduate students under the leadership of professor Leonard Kleinrock sent the first packet-switched message between two computers. A member of Kleinrock’s team, Charley Kline, had the distinction of being first to send it, but it was not a rousing start. As Kline at UCLA tried logging into the Stanford Research Institute’s computer for the first time, the system crashed just as he was typing the letter “G” in “LOGIN.”

The bugs were worked out, and further connections were made flawlessly, but the early network had many limitations. At the time of Kline’s first message to Stanford, logging into a remote computer was one of just three tasks possible on ARPANET; the other options were printing to a remote printer and transferring files between computers. Nevertheless, the interest generated by the nascent two-node network was intense. By the end of 1969, academic institutions were scrambling to connect to ARPANET. The University of California–Santa Barbara and the University of Utah linked up that year. By April 1971, there were 15 nodes and 23 host terminals in the network. In addition to the four initial schools, contractor BBN had joined, along with MIT, the RAND Corporation, and NASA, among others. By January 1973, there were 35 connected nodes; by 1976, there were 63 connected hosts.

During its first 10 years, ARPANET was a test bed for networking innovations. New applications and protocols like Telnet, file transfer protocol (FTP), and network control protocol (NCP) were constantly being devised, tested, and deployed on the network. In 1971, BBN’s Ray Tomlinson wrote the first e-mail program, and the ARPANET community took to it instantly. “Mailing lists,” which eventually became known as “LISTSERVs,” followed e-mail almost immediately, creating virtual discussion groups. One of the first e-mail discussion lists was SF-LOVERS, which was dedicated to science fiction fans.

What ARPANET could not do was talk to any of the other computing networks that inevitably sprang up in its wake. Its design required too much control and too much standardization among machines and equipment on the network, according to Naughton. So in the spring of 1973, Vinton Cerf and Bob Kahn began considering ways of connecting ARPANET with two other networks that had emerged, specifically SATNET (satellite networking) and a Hawaii-based packet radio system called ALOHANET. One day, waiting in a hotel lobby, Cerf dreamed up a new computer communications protocol, a gateway between networks, which eventually became known as transmission-control protocol/Internet protocol (TCP/IP). TCP/IP, which was first tested on ARPANET in 1977, was a way that one network could hand off data packets to another, then another, and another. Eventually, when the Internet consisted of a network of networks, Cerf’s innovation would prove invaluable. It remains the basis of the modern Internet.

In 1975, ARPANET was transferred to the Defense Communications Agency. By that time, it was no longer experimental, nor was it alone. Numerous new networks had emerged by the late 1970s, including CSNET (Computer Science Research Network), CDnet (Canadian Network), BITNET (Because It’s Time Network), and NSFNET (National Science Foundation Network); the last of these would eventually replace ARPANET as the backbone of the Internet, before it was itself superceded by commercial networks.

The term “Internet” was adopted in 1983, at about the same time that TCP/IP came into wide use. In 1983, ARPANET was divided into two parts, MILNET, to be used by military and defense agencies, and a civilian version of ARPANET. The word “Internet” was initially coined as an easy way to refer to the combination of these two networks, to their “internetworking.”

The end of ARPANET’s days arrived in mid-1982, when its communications protocol, NCP, was turned off for a day, allowing only network sites that had switched to Cerf’s TCP/IP language to communicate. On January 1, 1983, NCP was consigned to history, and TCP/IP began its rise as the universal protocol. The final breakthrough for TCP/IP came in 1985, when it was built into a version of the UNIX operating system. That eventually put it in Sun Microsystems workstations and, Naughton writes, “into the heart of the operating system which drove most of the computers on which the Internet would eventually run.” As Cerf would observe, “The history of the Net is the history of protocols.”

As both free and commercial online services like Prodigy, FidoNet, Usenet, Gopher, and many others rose, and as NSFNET became the Internet’s backbone, ARPANET’s importance diminished. The system was finally shut down in 1989 and formally decommissioned in 1990, just two years before Tim Berners-Lee would change everything all over again with the introduction of the World Wide Web.