Internet Guide Logo


bullet Introduction

ARPANET (Advanced Research Projects Agency Network) was one of the first 'wide area' computer networks, and was one of the first computer networks to feature packet switching. Packet switching underpins how the Internet works, and ARPANET is viewed as a forerunner to the Internet. ARPANET allowed the U.S. military, research laboratories and universities to connect, communicate and exchange data. ARPANET is a term that can be broken down into two segments (ARPA and NET): ARPA stands for the Advanced Research Projects Agency, and NET stands for network.

bullet The Plan to build ARPANET

In the early 1960s, one of the first scientists to formulate the idea of an 'Intergalactic' (global) computer network, using a standardised computer language, was Joseph "Lick" Licklider. Licklider was the first director of ARPA's Information Processing Techniques Office (IPTO), and served as director from 1962-1964. ARPA was an agency within the United States Department of Defense (Government department), and was eventually renamed to DARPA. Licklider was somewhat of an evangelist in promoting the idea of a wide area network. Licklider was not involved in the development of ARPANET, but his influence was crucial in persuading Ivan Sutherland and Robert Taylor (his successors) to build the network.

Joseph Licklider
(Pictured: Joseph Licklider)

In 1964, Licklider was replaced as the director of the Information Processing Techniques Office (IPTO) by Ivan Sutherland, and in early 1965, Sutherland provided the funding for an early experiment in networking. The experiment was conducted by Lawrence Roberts and Thomas Marill, and the result proved a success: it connected a TX-2 computer to a Q-32 mainframe computer. The TX-2 computer was located at Roberts MIT Lincoln Lab (east coast of America) and the Q-32 mainframe computer was located at Marill's SDC's Lab (west coast of America). Licklider had previously met Lawrence Roberts at a Homestead Meeting, and had evangelized Roberts about the idea of a global communications network.

In 1966, Sutherland was replaced as the director of the IPTO by Robert 'Bob' Taylor, who was impressed by Roberts and Marill's network experiment, and wrote an early plan for a 'wide area' network. In 1966, Bob Taylor received $1,000.000 from Charles Herzfeld (director of ARPA) to build a wide area computer network, and Lawrence 'Larry' Roberts was persuaded to leave MIT's Lincoln Laboratory to build ARPA's computer network. Initially, Lawrence Roberts refused the offer, as he was happy working at MIT's Lincoln Lab, but pressure was applied by Herzfeld - ARPA was providing half of the funding for the Lincoln Lab - and the director of the Lincoln Lab explained to Roberts that it may be a good career decision to move to ARPA.

In 1967, Lawrence Roberts became the chief scientist of the ARPANET project and is credited as it's architect. Plans to build ARPANET began in 1967, and early plans were discussed and outlined by Roberts at the ARPANET Design Session; which were held at Ann Arbor, Michigan. Roberts outlined the idea of a network with a central mainframe computer providing the networking function; an idea that proved unpopular. Wes Clark slipped a note to Roberts stating "you've got the network inside out", Clark believed the power of the network should be located at the edges and not the center, and explained to Roberts that the networking function should be handled be specialised minicomputers.

The specialised minicomputers would speak a universal network language, and would ensure that the computers that connect to the minicomputers would only need to translate the universal network data. Clark's idea simplified the network and impressed Roberts, who outlined the concept of the specialised minicomputers in a paper titled: Multiple Computer Networks and Intercomputer Communications. Barry Wessler assisted Larry Roberts in finalising the specification for the minicomputers (named IMPs) and E. B. Shapiro, who worked at the Stanford Research Institute, authored a report titled A Study of Computer Network Design Parameters which aided Wessler and Roberts in defining their specification for the IMPs.

By 1968, the plans for ARPANET were well under way, and Lawrence Roberts contracted Leonard Kleinrock at UCLA to undertake measurement calculations for the network. Roberts sent a Request For Quotation to over 100 companies to build the IMPs (minicomputers) for ARPANET. ARPA signed a contract with Bolt, Beranek and Newman (BBN) technologies to build the IMPs, and Frank Heart's was the head of the BBN group that built them. The Interface Message Processor (IMP) was a micro-computer that functioned as a packet-switching node, and the IMP technology was outlined in the first RFC document (RFC 1). Robert Kahn, who was part of Frank Heart's BBN group, authored Host to IMP Spec. 1822, a paper that outlined how to write an interface for connecting host computers to IMPs.

Leonard Kleinrock standing next to ARPANET's IMP Interface Message Processor
(Pictured: Leonard Kleinrock standing next to ARPANET's Interface Message Processor)

BBN IMP Team: T.Thatch, B.Bartell, D.Walden, J.Geisman, R.Kahn, F.Heart, B.Barker, M.Thorpe, W.Crowther, and S.Ornstein
(Pictured: BBN IMP Team: T.Thatch, B.Bartell, D.Walden, J.Geisman, R.Kahn, F.Heart, B.Barker, M.Thorpe, W.Crowther, and S.Ornstein)

The task of writing the host-to-host protocols for ARPANET was conducted by the Network Working Group (NWG), which was led by Stephen D. Crocker and supervised by Larry Roberts. It was Crocker who invented the Request for Comments (RFC) document system, and wrote the earliest information about ARPANET protocols in RFC1, RFC2, RFC3 and RFC4. As previously stated, BBN had provided the '1822 protocol', outlined in BBN Report 1822, and ARPANET's host-to-host protocol program was designed using this documentation. The Network Control Program (NCP) was the network protocol designed by the NWG, which was a simplex protocol that utilised two ports, and was similar to the role transport layer protocols perform in TCP/IP.

Stephen Crocker stated that the process of designing the first network protocols for APRANET was a haphazard process that did not adhere to a 'grand plan'. On the 3rd of June 1969, the official plan for ARPANET was delivered to Bob Taylor, and the scientists who designed and built ARPANET are now described as the Pioneers of the Internet.

bullet ARPANET is operational

The first IMP shipped by Bolt, Beranek and Newman technologies was delivered to the University of California, Los Angeles (UCLA) in August 1969. The team involved in its installation included: Vint Cerf, Bill Naylor, Jon Postel, Steve Crocker, and Mike Wingfield. By September 1969, a message had been sent between UCLA's IMP and their mainframe computer (SDS Sigma 7), making it the first ARPANET node. By October 1969, the first connection of two nodes was conducted: the first node was located at UCLA and the second node was located at the Stanford Research Institute (SRI). Leonard Kleinrock and Douglas Engelbart, respectively, led the teams who conducted this test. The first successful message sent on ARPANET was "login" and it was transmitted by Charles S. Kline at 10:30pm on the 29th of October 1969.

When ARPANET was eventually launched, it consisted of four nodes, with each node featuring an Interface Message Processor (IMP). The four nodes were located at:

  1. University of California, Los Angeles (UCLA) - (SDS Sigma 7 computer) (Sign EXtend software)
  2. Stanford Research Institute (SRI) - (SDS 940 computer) (NLS software)
  3. University of California, Santa Barbara (UCSB) - (IBM 360/75 computer) (OS/MVT software)
  4. University of Utah - (DEC PDP-10 computer) (TENEX software)

The first test of ARPANET proved a success and the importance of ARPANET led to Lawrence Roberts becoming the new director of the IPTO. In 1969, Bob Taylor had been sent to probe conflicting reports coming from the Vietnam War, and could no longer oversee the ongoing development of ARPANET. Lawrence Roberts was director of the IPTO from 1969-1973, and was responsible for expanding the network across North America and internationally.

The first authority created for ARPANET was the Network Information Center (NIC), the NIC was located at Douglas Engelbart's Augmentation Research Center (SRI) laboratory, and was created to manage the process of assigning hostnames. Hostnames were devised to make it easy to memorise the address of ARPANET hosts. ARPANET nodes could request a hostname from the Network Information Center (NIC), and the NIC would manually store the hostname information in a file named: HOSTS.TXT. Nodes would download this file from SRI and would manually install it. Eventually, the Domain Name System (DNS) was created (1984) to replace the system managed by the NIC.

By 1973, Lawrence Roberts had overseen the connection of ARPANET to over 25 nodes in North America, and its first international nodes (in Norway and England). In 1973, the Mansfield Amendment directed ARPA to spend less funding on science projects and to focus on projects with a direct military use. While ARPANET would continue to be funded, it's funding was limited, and many of the computer scientists who created it moved to the private sector to develop commercial computer products; Lawrence Roberts was amongst these scientists. In 1975, due to the Mansfield Amendment, the Defense Communication Agency (DCA) took operational control of ARPANET, and the DCA split ARPANET into a research and a military network. The military network was named MILNET and the research network was named ARPANET. Longterm plans were to phase-out the research ARPANET network.

Despite the turbulence surrounding the operational control of ARPANET, the development of computer networking protocols was still funded by DARPA (previously named ARPA). Bob Kahn joined DARPA in 1972, and set about designing an 'open' network architecture that would solve the drawbacks of the NCP program. In 1973, Bob Kahn invited Vint Cerf to help him develop his idea into a working protocol: this protocol took nearly a decade to be standardised and was eventually named TCP/IP. On the 1st of January 1983, TCP/IP replaced the Network Control Program (NCP) on ARPANET, a date that is often referred to as being the day the Internet was born.

ARPANET was slowly decommissioned from 1985 to 1989: due to the creation of NSFNET and the Federal Internet Exchange (FIX). The FIX allowed federal networks, like MILNET, to connect to the NSFNET backbone.