ARPANET (1969): The First Message and Birth of the Internet

The ARPANET was the first large computer network to use packet switching, and it is the direct ancestor of today’s internet. On the evening of 29 October 1969 a computer at the University of California, Los Angeles (UCLA) sent its first message to a computer at Stanford Research Institute (SRI) some 350 miles (about 560 kilometres) away. The message was only two letters long, but the network it travelled on grew over the next two decades into the global system that now links billions of devices.

Background: computers, the Cold War and ARPA

The Advanced Research Projects Agency (ARPA) was set up by the US Department of Defense in February 1958, a few months after the Soviet Union shocked the West with the launch of Sputnik 1. Its task was to fund ambitious research so that the United States would not be technologically surprised again.

Computers in the 1960s were large, expensive machines, each with its own software and way of working. ARPA paid for computers at many universities and research centres, and they could not easily share programs or data. In 1962 the psychologist and computer scientist J. C. R. Licklider became head of ARPA’s Information Processing Techniques Office (IPTO). He wrote memos describing an “Intergalactic Computer Network”, a half-humorous name for the idea of linking computers so that people could use them together.

In 1966 Robert Taylor, then director of IPTO, had three separate terminals in his office, each connected to a different remote computer. Frustrated, he persuaded ARPA’s director, Charles Herzfeld, to fund an experimental network. Herzfeld agreed to an initial budget of around one million dollars in a famously short meeting.

Packet switching

The ARPANET relied on a new idea called packet switching. Instead of opening a dedicated circuit between two machines, as a telephone call does, a message is cut into small blocks, or packets. Each packet carries an address and travels through the network independently, and the pieces are put back together at the destination.

The concept was developed independently by several people. Paul Baran at the RAND Corporation described a distributed network of message blocks in the early 1960s, in studies on communications that could survive an attack. In Britain, Donald Davies at the National Physical Laboratory reached similar ideas in 1965 and coined the word “packet”. At UCLA, Leonard Kleinrock had studied the mathematics of queues and data flow in networks in his doctoral work in the early 1960s.

Taylor hired Lawrence (Larry) Roberts, a computer scientist from MIT’s Lincoln Laboratory, to manage the project. In 1967 Wesley Clark suggested that each research computer, or host, should not handle the network itself. Instead, a small dedicated computer at each site would route the packets. These machines were called Interface Message Processors, or IMPs, and they were the first routers.

Building the network

ARPA issued a request for proposals in the summer of 1968. At the end of that year the contract to build the IMPs went to Bolt Beranek and Newman (BBN), a consulting firm in Cambridge, Massachusetts. A small team led by Frank Heart, including Robert Kahn, Severo Ornstein, Will Crowther and Dave Walden, adapted a ruggedised Honeywell DDP-516 minicomputer for the job.

The sites themselves had to agree how their computers would talk to each other. Graduate students at the first sites formed the Network Working Group. On 7 April 1969 Steve Crocker of UCLA circulated its first note, called Request for Comments 1. The RFC series, intended as informal working papers, still records the technical standards of the internet today. Its modest, open style, inviting anyone to comment and improve on proposals, became a lasting part of the internet’s culture.

The first IMP was delivered to Kleinrock’s laboratory at UCLA at the end of August 1969 and was connected to the university’s SDS Sigma 7 computer in early September. The second arrived in October at SRI in Menlo Park, where it was linked to an SDS 940 used by Douglas Engelbart’s research group. The sites were joined by telephone lines leased from AT&T running at 50 kilobits per second.

The first message: “LO”

On the evening of 29 October 1969 Charley Kline, a student programmer working under Kleinrock at UCLA, sat at a terminal to log in to the SRI computer. At SRI, the programmer Bill Duvall was on the telephone with him to confirm each character as it arrived.

The plan was simple. Kline would type the word LOGIN, and the SRI computer was programmed to complete the command automatically once it recognised the first letters. Kline typed “L”, and SRI confirmed it. He typed “O”, and SRI confirmed it. When he typed “G”, the SRI system crashed. So the first message ever sent across the ARPANET was “LO”. About an hour later, after Duvall had fixed the problem, the full login succeeded. A handwritten entry in UCLA’s IMP log, timed at 22:30, simply records that the team had talked to SRI host to host.

Kleinrock later joked that “lo”, as in “lo and behold”, made a fitting first word. The exchange was not publicised at the time. Only years later, as the internet grew, did the date come to be seen as a milestone.

Growth in the 1970s

By December 1969 the network had four nodes: UCLA, SRI, the University of California at Santa Barbara and the University of Utah. More universities, government laboratories and defence contractors joined over the following years. By 1971 the Network Control Program, the first full host-to-host protocol, was in use across the network.

In 1971 Ray Tomlinson at BBN sent the first network email between two computers and chose the @ sign to separate user names from machine names. Email quickly became the network’s most popular use, far beyond what its designers had expected. Other early services included remote login, known as Telnet, and the File Transfer Protocol (FTP) for moving files between distant machines, both of which remained in wide use for decades.

In October 1972 Robert Kahn organised a large public demonstration of the ARPANET at the International Conference on Computer Communication in Washington, D.C., which convinced many engineers that packet switching worked. In 1973 the network made its first international links, to the NORSAR seismic array in Norway and to University College London.

From ARPANET to the internet

Other packet networks were appearing, using radio, satellite links and different hardware. Kahn, who had moved to ARPA, and Vinton Cerf, then at Stanford, set out to connect them. In May 1974 they published a paper describing a Transmission Control Program, which allowed different networks to exchange packets through gateways. The design was later split into two parts, TCP and IP, and the joined-up system came to be called the internet.

On 1 January 1983, often called “flag day”, all hosts on the ARPANET switched from the old protocol to TCP/IP. The same year, military sites moved to a separate network called MILNET. In the mid-1980s the National Science Foundation built a faster academic backbone, NSFNET, and it gradually took over the ARPANET’s role. The ARPANET was formally decommissioned in 1990.

That same period saw Tim Berners-Lee at CERN propose the World Wide Web, a system of linked documents that ran on top of the internet and made it accessible to the general public in the 1990s.

Myths and facts

A common claim is that the ARPANET was built to keep military communications running after a nuclear war. In fact its main purpose was to let ARPA-funded researchers share computing resources. Baran’s work at RAND had been motivated by survivability, and robustness was a desirable feature, but Taylor, Herzfeld and others repeatedly stressed that the ARPANET itself was a research network.

Another misconception is that the ARPANET and the internet are the same thing. The ARPANET was one network. The internet is a network of networks, made possible by TCP/IP, which the ARPANET adopted in 1983 and which outlived it.

By the numbers

Timeline

• 1958, February: The US Department of Defense creates ARPA.
• 1966: Robert Taylor wins funding for an experimental computer network.
• 1969, 7 April: Steve Crocker issues Request for Comments 1.
• 1969, August–September: The first IMP is installed at UCLA.
• 1969, 29 October: UCLA sends the first message, “LO”, to SRI.
• 1969, December: The network links four sites.
• 1972, October: Public demonstration of the ARPANET in Washington, D.C.
• 1974, May: Cerf and Kahn publish their design for internetworking.
• 1983, 1 January: The ARPANET switches to TCP/IP.
• 1990: The ARPANET is decommissioned.

Why it matters

The ARPANET showed that packet switching could link very different computers into a single working network. Its engineers also created the open, cooperative way of setting standards through RFCs. The first faltering “LO” in 1969 marks the starting point of a technology that has transformed communication, business, science and everyday life across the world.