THE ORIGINS OF THE ARPANET

The ARPANET was a packet-switched computer network commissioned by the
Advanced Research Projects Agency, an arm of the United States Department
of Defense. It is widely regarded as the technical precursor of the modern
Internet, and its development between 1966 and 1972 produced many of the
ideas — packet switching, distributed routing, host-to-host protocols, and
end-to-end design — that subsequent network designers built upon.

The intellectual origins of the project are usually traced to the work of
Joseph Carl Robnett Licklider. In 1962, while at Bolt, Beranek and Newman
in Cambridge, Massachusetts, Licklider wrote a series of memoranda
describing what he called an "Intergalactic Computer Network." Later that
year he became the director of the Information Processing Techniques Office
at ARPA, where he funded foundational work in time-sharing and human-machine
interaction at MIT, Stanford Research Institute, the University of
California at Los Angeles, and Carnegie Mellon. Licklider stepped down in
1964 and was succeeded by Ivan Sutherland and then by Robert Taylor. Taylor,
who took office in 1966, observed that ARPA-funded research labs across the
country each had their own incompatible terminals and accounting systems,
and proposed funding a single shared network to connect them.

Taylor brought in Lawrence Roberts, then at MIT Lincoln Laboratory, to lead
the project. Roberts' design called for a packet-switched network — an idea
that had been independently developed by Paul Baran at the RAND Corporation
in the early 1960s and by Donald Davies at the National Physical Laboratory
in the United Kingdom. Baran's 1964 reports described a "distributed adaptive
message block network" intended to survive partial destruction in a nuclear
exchange; Davies coined the term "packet" in 1965 and built a small network
at NPL to demonstrate the technique. Roberts incorporated both lines of
work into the ARPANET design, with Davies as a regular correspondent.

The contract to build the network's switching equipment was awarded to
Bolt, Beranek and Newman — Licklider's former employer — in December 1968.
The BBN team, led by Frank Heart and including Robert Kahn, Severo Ornstein,
and David Walden, designed and built the Interface Message Processor,
known as the IMP. The IMP was a Honeywell DDP-516 minicomputer modified to
run BBN's networking software; it sat between a host computer and the
network's leased telephone lines and handled packet routing on the host's
behalf. The first IMP was delivered to UCLA on August 30, 1969.

The first ARPANET node was installed at the University of California at
Los Angeles, in the laboratory of Leonard Kleinrock. Kleinrock had developed
the queueing theory underlying packet networks in his 1962 MIT doctoral
dissertation. The second node was installed at Stanford Research Institute
in Menlo Park, in the laboratory of Douglas Engelbart, whose NLS system
hosted some of the earliest networked applications. On October 29, 1969, a
graduate student named Charley Kline at UCLA attempted to log in to the
SRI host. The system crashed after he had typed the letters "L" and "O";
Kleinrock and others later considered "lo" — the first two letters of
"login" — to be the network's first transmitted message. By the end of
1969, four nodes were connected: UCLA, SRI, the University of California
at Santa Barbara, and the University of Utah.

Host-to-host communication required a protocol that was independent of any
particular computer's operating system. The Network Working Group, an
informal body convened by Steve Crocker at UCLA in 1968, produced what
became known as the Network Control Program, or NCP, in 1970. The NWG also
introduced the Request for Comments document series, which began in April
1969 with Crocker's RFC 1, "Host Software." The RFC series persists to this
day and is now maintained by the Internet Engineering Task Force; it is
widely credited with establishing an open, consensus-driven culture for
network standards.

By 1971 the network had grown to fifteen nodes and supported applications
including remote login (Telnet) and file transfer (FTP, originally specified
by Abhay Bhushan in RFC 114). Electronic mail emerged unplanned: Ray
Tomlinson, working at BBN in 1971, modified the existing SNDMSG and CPYNET
programs to send messages between hosts and chose the at-sign character
as a separator between user and host names. By the mid-1970s, electronic
mail was responsible for the majority of the ARPANET's traffic.

The shortcomings of NCP — its dependence on a reliable network and its
inability to interconnect with other networks — drove the next major
research effort. In 1973, Robert Kahn, by then at ARPA, and Vinton Cerf,
who had moved from UCLA to Stanford, began designing what they initially
called the Transmission Control Program. Their May 1974 paper, "A Protocol
for Packet Network Intercommunication," described a layered model in which
gateways interconnected diverse networks and the end hosts handled
reliability. The protocol was later split into two layers, TCP for
reliable transport and IP for routing, and was deployed on the ARPANET in
1983 in a transition known as flag day, displacing NCP entirely.

ARPANET service ended in 1990. By that time, the National Science
Foundation's NSFNET had taken over the academic backbone role, and
commercial Internet service providers were beginning to operate on the
broader IP network. The technical artifacts of the project — the IMP, the
NCP and TCP/IP protocols, the RFC document series, the host-to-host design
philosophy — outlived the network itself and shaped every subsequent
generation of Internet infrastructure.

Several of the project's early participants went on to receive the Turing
Award for their work on the network. Robert Kahn and Vinton Cerf shared the
2004 award for the design of TCP/IP. Leslie Lamport, who later worked on
distributed systems theory at SRI and elsewhere, received the 2013 award
for his foundational contributions to that field. The IMP itself, decorated
with hand-drawn lightning bolts by the BBN engineers, is preserved at the
Computer History Museum in Mountain View, California.
