ARPANET 1969-1990 Computer Network The Advanced Research Projects Agency Computer Network, known as the ARPANET, was the first large-scale interconnection of dissimilar mainframe computers. It demonstrated that packet switching of data was more appropriate for computer to computer communication than telephone-like circuit switching and it established a tradition of collaborative computer science research and open documentation that set the basis for the Internet. The ARPANET began at the end of 1969 as a four node experiment to achieve resource sharing and to serve as a test bed for computer communication research. That beginning came after six years of active preparation by the computer science community. In 1962, the ARPA director brought J.C.R. Licklider in to set up an Information Processing Techniques Office (IPTO) with which to encourage and support advanced computer research. Licklider took support of experimentation in interactive computing as a first step toward what he envisioned as the "Intergalactic Network." This network would consist of the connection of all computers and computer users regardless of geographic location. By 1965, IPTO awarded a contract for a cross country connection between Lincoln Laboratory's TX-2 computer in Massachusetts and a System Development Corporation's AN/FSQ-32 computer in California. This experiment was successful in demonstrating that long distance linking of computers was possible but that telephone switching was problematic. Research in the UK led by Donald Davies suggested that a system based on data packets was an alternative to telephone switching for computer data. Work by Leonard Kleinrock on the mathematic theory of queueing suggested the same alternative. At the Principal Investigators meeting in 1967, a gathering of the directors of IPTO projects at the University of Michigan, Larry Roberts presented a proposal to finalize specifications for a larger scale experimental network. Wes Clark proposed that a smaller computer (IMP, or Interface Message Processor) be inserted between each mainframe computer (host) and the telephone lines. Based on the collaboration of the IPTO researchers, a Request for Quotation (RFQ) to build the IMPs was released in July of 1968. IPTO did not want its research stifled by financial pressures so it budgeted $500,000 in 1968 for this networking experiment. At about the same time IPTO gathered a meeting of computer science graduate students from what would be the first four host sites so they too would be involved from the start. This meeting in the Summer of 1968 began the collaboration among these students to solve the fundamental host-to-host problems. Out of this meeting came the Network Working Group (NWG). The students wondered if some expert would come and give direction to the work. Instead they began to record their thoughts and deliberations in a set of notes Steve Crocker named Request for Comment (RFC). The RFCs were open to all who wanted to contribute. Crocker wrote, "The content of a NWG note may be any thought, suggestion, etc related to the HOST software or other aspect of the network." Via their meetings and RFCs the students worked out what became the Network Control Program or Protocol (NCP) and the first services on the ARPANET: terminal access to remote hosts (Telnet), file transfer (FTP) and remote job entry (RJE). In December 1968, an engineering consulting firm, Bolt Beranek and Newman (BBN), was awarded the contract to configure and deploy the Interface Message Processors and to control the inter- IMP communications. AT&T was to provide dedicated long distance telephone lines capable of transporting 50,000 (50K) bits of data per second to link the IMPs. BBN was also to coordinate with the graduate students the linking of each host to its individual IMP. Although the RFQ contained the basic design for the ARPANET, BBN's team including Frank Heart, Servo Ornstein, Robert Kahn, Dave Walden, Will Crowther and others had a large number of conceptual and technical problems to overcome to meet their deadline. The RFQ required BBN to deliver one IMP per month beginning on September 1, 1969. BBN's team succeeded in creating software programs that would automatically keep the network functioning on the chosen IMP computer, Honeywell DDP 516. Also, the team created the interface cards that contained the electronics necessary for each IMP to communicate in one direction with a host and in the other direction across the leased phone lines with another IMP. In this way the IMPs and telephone lines could become a communications subsystem which would be uniform, under the central control of BBN and physically separate from the hosts. If a host was to fail it would not disturb the communications subnetwork and if the subnetwork were to fail it would not impair the operation of any host. To the surprise of the researchers in California, BBN delivered the first IMP on time to the University of California at Los Angeles. By the end of 1969, BBN had also delivered IMPs to the University of California at Santa Barbara, the Stanford Research Institute and the University of Utah. And thus was born the ARPANET. BBN was able to conduct communication experimentation on this four-node network while it continued to deliver one IMP per month to new IPTO contractor sites that IPTO chose to attach hosts to the network. By June 1970 there were two cross country lines connecting MIT, BBN and Harvard with six West Coast IMPs. The BBN team and other researchers had many new communication problems to solve. Packet switching had to be designed from scratch. In telephony, first a path is established and then the whole call occurs over that circuit. In packet switching each small chunk of data called a packet carries the destination address with it and gets routed at each IMP toward the destination. So the researchers had to work out mechanisms for addressing and routing of those packets. They had to find ways of controlling packet flow and congestion. And they had to build in automatic error detection and correction. The network itself was a great help because it allowed the trial of new solutions and others to give feedback to suggest improvements. The BBN team and IPTO community succeeded in overcoming each false start. For example, the original routing procedure was for each IMP to automatically update its neighboring IMPs with its idea of the state of the network. Some packets looped back and forth in the network without reaching their destination. The problem was the updates were too frequent. The solution was less frequent updates. The addressing scheme originally chosen allowed for a maximum of 64 IMPs and at most 4 hosts at each IMP. As the ARPANET succeeded in growing it was realized that the original thinking was not optimistic enough. While the packet switching technology was being worked out, the graduate students worked toward solving their basic problems: how to get hosts with different operating systems, word sizes, file systems, character sets, etc. to communicate with each other. Their solution was an agreement on a common set of communications conventions that came to be called "protocols". The process lead to a bakeoff in October 1971 at MIT. There, representatives from the first four sites attempted to log in to all of each other's sites. In so doing they collectively saw the problems and solutions and finally agreed on the specifications that became known as the host-to-host protocol NCP. Thus began the possibility of the use of the network for other than testing purposes. But what to use the network for? The first major use of the ARPANET was unplanned. In 1971 two researchers composed messages to each other as test data and so was started email. By using the network to leave a file with a message or question on someone else's host computer a rapid inexpensive form of human to human communication was invented. Other general uses were harder to find until Bob Kahn started to push each site to have something desirable to do. He conceived of a public demonstration of the ARPANET to take place in October 1972 at an international computer communications conference. The result was a great success including 40 terminals serving 1000 attendees who could log in and view for example a distributed air traffic control simulation. At this conference the reality of computer networking became clearer and researchers from around the world went home anxious to develop similar networks. They also set up the International Network Working Group (INWG) which was to play a role in solving the problem of linking all the different networks that were going to appear. In 1973 an IMP was installed in Hawaii and the mainframe computer at the University of Hawaii was added to the ARPANET by using radio waves and a satellite rather than wires to transport packets. Soon an IMP in London was also connected via satellite. Also in 1973 in a separate but related development, Robert Metcalfe was inventing the Ethernet method of connecting nearby computers with each other via a single cable. These events showed that the ARPANET was just the beginning of computer networking. Between 1973 and 1975 the ARPANET expanded from 30 to 57 nodes, and evolved from an experimental to more of an operational network. A study was commissioned to determine who would own and manage the network. The resulting proposal that the ARPANET be spun off to a common carrier like AT&T was rejected. A government auditor had worried out loud about the propriety if a government- developed network would be given to a common carrier which would then make profit selling network service back to the government. Instead from July 1 to December 31, 1975 administration of the ARPANET was transferred to the Defense Communication Agency (DCA). In 1975 there was experimentation by IPTO researchers linking the ARPANET pairwise to PRNET, a ground radio network and SATNET, a satellite network. Also in that year the MsgGroup mailing list was started to explore the use of the ARPANET for conferencing. Mailing lists were uses of email where each message sent to the list was distributed to all list members. They each could read it and comment by sending a new message which in turn would be sent to all the other list members. Very serious and high level results occur. Other mailing lists appeared such as Human-nets for the discussion of the future of ubiquitous networking and SF-Lovers for the discussion of a favorite hobby, science fiction. Participants in these lists were happily surprised how interesting and valuable such network use was. The late 1970s saw intensified research to determine how the ARPANET and other networks could participate in a broader sharing of resources. The TCP/IP protocols emerged as a way forward toward such an internetworking. In 1979, the Usenet newsgroup system was invented by graduate students Tom Truscott and Jim Ellis. It was seen as a "poor man's ARPANET" because it allowed for networking among people who where not at ARPA/IPTO supported sites. When Mark Horton put ARPANET mailing list messages into Usenet newsgroups and then vice versa, the Usenet users were poorly received. But the mixing of the two communities eventually enhanced the value of both. By the early 1980s the Department of Defense decided to require that all its nodes use TCP/IP by January 1, 1983. The transition was difficult because it required a coordinated change at many separate sites and was not quite completed on time but was successful. Ten months later, the approximately 140 nodes serving 320 computers of the ARPANET were divided into two separate networks connected by a gateway. The resulting network connecting the computers that were used for scientific and educational purposes retained the name ARPANET. The other smaller network to serve unclassified defense needs was named MILNET. This split of the original ARPANET into two linked networks was the embryo of the Internet. Demand for network connection grew in the 1980s. The ARPANET served as a stimulus for the formation first of Computer Science Network (CSNET) and then the National Science Foundation Network (NSFNET). As Local Area Networks (LANs) grew in number, the ARPANET began to play more of a backbone role of interconnecting other networks. As the NSFNET took on more of this role the ARPANET was phased out. The ARPANET was decommissioned in 1990 after 30 years of experimentation and service. In those years the ARPANET dramatically demonstrated the feasibility and efficiency of packet switching communication, the desirability and productivity of resource sharing and the value of open standards and collaborative research and development. The great success of computer communications owes a great deal to the vision and scientific and engineering skill of the ARPANET pioneers. References Hafner, Katie and Matthew Lyon. Where Wizards Stay Up Late: The Origins of the Internet. New York: Simon and Schuster, 1996. Hauben, Michael and Ronda Hauben. Netizens: On the History and Impact of Usenet and the Internet. Los Alamitos, CA.: IEEE Computer Society Press, 1997. Lynch, Daniel C. and Marshall T. Rose. Internet System Handbook. Reading, MA.: Addison-Wesley, 1993. ------- Suggested Graphics A map of the ARPANET in December 1969 for example on page III-97 of the ARPANET Completion Report by Heart, McKenzie, McQuillian and Walden, BBN 1978. Also on page 352 of McKenzie, Alexander A. and David Walden. "ARPANET, the Defense Data Network and Internet." In The Encyclopedia of Telecommunications. Vol 1. Edited by Fritz E, Froehlich, Allen Kent and Carolyn M. Hall. 341-376. New York. Marcel Dekker, 1991. A map of the ARPANET in July 1982 (or later) for example on page 368 of McKenzie and Walden (see above) A photograph of an IMP for example Hafner and Lyons between pages 160 and 161 (see references above).