The Off-Key Bard
History

This Day in History — ARPANET Sends Its First Message

On this day in 1969

On October 29, 1969, UCLA's Charley Kline tried to type LOGIN to a computer 400 miles away at Stanford. The system crashed after two letters: LO.

The Off-Key Bard enters Room 3420 of Boelter Hall at the University of California, Los Angeles, on an October evening in 1969. The room holds enormous computers, cables and switches that look nothing like the devices we use today. There are no smartphones, no web browsers, no social media, no Google. Two programmers, separated by hundreds of miles, are about to connect their computers through an experimental network. The word they intend to send is simple: LOGIN.

They manage only two letters before the system crashes. LO.

On October 29, 1969, a UCLA programmer named Charley Kline sent the first host-to-host message over the ARPANET, an experimental network funded by the U.S. Defense Department's Advanced Research Projects Agency. The message went from UCLA to the Stanford Research Institute (SRI) in Menlo Park, about 400 miles away, where Bill Duvall was at the other end. Kline typed the letters of LOGIN one at a time. The first two arrived. The receiving system crashed on the third.

Computers That Could Not Talk

In the 1960s computers were expensive, room-sized and mostly incompatible with one another. A researcher who needed a particular machine often had to go where it was, and data moved between institutions on tape, cards and paper.

The people who wanted to change this had a vision before they had the means. J. C. R. Licklider, writing at ARPA in 1963, described a distributed network of computers. ARPA itself had been created on February 7, 1958, under President Eisenhower, after the Soviet Union's launch of Sputnik 1 in 1957. In 1967 Lawrence Roberts, who was running the ARPANET program, adopted packet switching after hearing the work of Donald Davies's team at a symposium.

Packets and the IMP

Telephone networks used circuit switching, which reserves a dedicated path for the whole of a call. Packet switching occupies a channel only while each packet is being sent, freeing it for other traffic. Computers talk in bursts, so the second approach suited them better. Several people contributed to the idea. Paul Baran at RAND proposed what he called distributed adaptive message block switching, Donald Davies at Britain's National Physical Laboratory independently developed a similar concept and coined the term packet switching, and Leonard Kleinrock at UCLA applied queueing theory to network delays. Kleinrock later claimed priority for the idea, which other pioneers disputed, and historians generally credit Baran and Davies with inventing digital packet switching.

To connect dissimilar computers without making each one handle the network, ARPA used a small dedicated computer at each site. Bolt, Beranek and Newman (BBN) built these Interface Message Processors, or IMPs, on ruggedized Honeywell DDP-516 minicomputers. The IMP is considered the first generation of what we now call routers. BBN delivered the first one to UCLA on August 30, 1969.

Four Sites

Roberts chose the first four sites because they were already ARPA sites: UCLA, where Kleinrock's team ran the Network Measurement Center, SRI, the University of California, Santa Barbara, and the University of Utah. The UCLA–SRI link used a 50-kilobit-per-second line leased from AT&T.

The Night of October 29

The test was meant to be simple. Kline, a student programmer on Kleinrock's team, would log in to the SRI computer from a terminal at UCLA. He and Duvall also had telephone headsets so they could talk during the test. Kline typed an "l" and asked whether it had arrived. It had. He typed an "o," and that arrived too. When he typed the "g," the system crashed.

Accounts of the timing differ slightly. The Computer History Museum places the first attempt at around 9 p.m., with the connection functioning by approximately 10:30 p.m. Leonard Kleinrock later explained that the crash resulted from a memory overflow on SRI's host computer rather than a failure of the ARPANET connection itself. Once the problem was corrected, the full login succeeded later that evening. The precise timing varies between historical accounts, but all agree on the essential sequence: the letters "L" and "O" were successfully transmitted, the receiving computer crashed, and the complete login was accomplished afterward.

The entry in the IMP log, written by Kline in Boelter Hall at 22:30, reads "Talked to SRI Host to Host." As Kleinrock put it, there was no camera or voice recorder, and the log entry is the only record. No one planned a famous first line. The first message was "lo" because that was as far as the login got, and it makes a fitting pun, as in lo and behold.

From Four Nodes to a Network

By the end of 1969 the first four sites were reported to be active, though the Computer History Museum's own account is inconsistent, with the main text placing UCSB and Utah in early 1970 and an addendum placing them by the end of 1969. The University of Utah was connected in December 1969.

The network soon carried more than logins. In 1971 Ray Tomlinson, an engineer at BBN, used the @ sign to separate a user's name from the name of the host computer in a system that sent messages between computers, and email became one of the network's most important uses. In October 1972 Robert Kahn of BBN organized a demonstration at the International Computer Communications Conference at the Washington Hilton. About 800 people attended, and forty terminals gave attendees access to a network that by then had 29 host sites.

A Network of Networks

ARPANET was only one network, and others used radio, satellite or telephone links. Vint Cerf and Robert Kahn's 1974 paper, "A Protocol for Packet Network Intercommunication," introduced TCP, which was later split so that IP was separate. In 1981 Jon Postel published a plan to move ARPANET's roughly 400 hosts from the older NCP protocol to TCP/IP, and hosts that had not switched by January 1, 1983 would be disconnected. Cerf considers that day the one on which "the operational Internet was born."

The Internet and the Web are not the same thing. Tim Berners-Lee developed the Web at CERN in 1989 and 1990, and it opened to the public in 1991. It runs on top of the Internet.

Then and Now

The original link ran at 50 kilobits per second. A one-gigabit connection is about 20,000 times faster. The principle is recognizably the same: information is divided into packets, networks forward them, and devices that follow compatible protocols reassemble them. Today's network also brings problems the researchers of 1969 were not thinking about, including security, privacy, misinformation and our dependence on a system we rarely see.

Room 3420 is still there. UCLA restored it to its 1969 appearance, with walls painted lime green to match, and it opened on October 29, 2011 as the Kleinrock Internet Heritage Site and Archive. It holds IMP No. 1, a tall machine weighing 800 pounds, which Kleinrock said should stay where it served its life.

Two letters crossed the miles that night,
Before a crash cut short their flight.
From "LO" a wider world would grow—
Connecting places we could not yet know.

Technological revolutions do not always begin with an announcement. Researchers identify a problem, engineers build a system, programmers test it, something breaks, and they repair it and try again.

On October 29, 1969, Kline and Duvall were not trying to invent email, online shopping or the Web. They were showing that two different computers could communicate through a shared packet-switched network. The original ARPANET is gone, but its ideas are not, and billions of people now rely on networks that grew from it.

Sources

← Back to the Desk — History