Architects of the Net of Nets - Part 4
by Dick Reiman, Historian
By 1972, the ARPA network had been accepted as a reliable large scale network. Robert Kahn and Vincent Cerf had together been the catalyst to its success. Bob Kahn, as a program manager at ARPA was leading advancements in satellite radio packet networks used by the military. The solution found to connecting radio networks was to be applicable to connecting any dissimilar network. This solution could then be applied to ARPA, and connections to network unrelated to ARPA such as a phone network.
Kahn asked Cerf, currently an electrical engineering professor at Sanford University, to collaborate; Cerf immediately agreed. The task was to link disparate networks. Kahn knew the networking and Cerf understood the programming protocols of the host computers, which connected the packet switching IMP to the site computer. Both knew enough of each other "turf" to discover flaws in the logic either proposed. A guiding principle was that no changes were to be needed in participating networks.
The solution was to place a gateway computer between networks. Now known as a "router", its job was to adjust arriving packets to the requirement of its network. This could mean repackaging incoming 500 bytes into two 250-bytes lengths, before routing them to their destination.
The basic design had to be simple. Considered were two basic categories: host and network functions. The protocol between hosts must not only enable end-to-end communication, but also to verify packet arrivals and prevent congestion by limiting arrival packet rates. Their two-part addressing scheme specified the network and individual host on the network. The first part told the sender's gateway the next gateway enroute to the destination network. Once inside the intended network, the host address was used to route the packet to the destination.
Kahn and Cerf had arrived at the definition of the Internet. Their paper on this subject, covering ideas for Transmission Control Protocol (TCP), was presented to the International Network Working Group at the University of Suffix, Brighton, England in September 1973. Later, TCP became TCP/IP. Kahn funded three contracts to implement the new design. One went to Cerf at Sanford, second to Ray Tomlinson at BBN, and to Peter Kirstein at University College, London.
The moment for testing was at hand with little media attention given to the event. On Nov. 22,1977, Cerf's crew demonstrated a triple-network Internet. Radio receivers dotted the hills around Menlo Park, so that a moving van with a packet radio terminal could send Internet packets into the ARPA's land lines and through satellites to Norway and University College, London. The packets were returned through the Atlantic packet satellite network to West Virginia back into the ARPA where they hopped to Machine C at UCLA's Information Science Institute in Los Angeles. The route crossed the Atlantic twice, went coast to coast, and didn't miss a beat. It proved that it was possible to drive down a highway and compute on machines in London or Los Angeles.
To the military, it demonstrated that mobile units in the field could tap resources in Arpanet in the U.S.