Early networking
Roberts' Arpanet program had an economic rationale. At that time, computers were multimillion-dollar mainframes and consequently precious resources. If this new "packet-switching" network could work, campuses doing Pentagon research would be better able to share computers.
The project had military appeal, too, as Paul Baran at RAND Corp. showed several years earlier. Because it called for decentralized, adaptive-routing capabilities, an enemy would have no central switching centers to target.
Circuit-switched telephone lines were fine for voice but distinctly uneconomical for bursts of data sent long intervals apart. Kahn drew an analogy with the highway system. Apply circuit-switching to the drive from Washington, D.C., to New York City, he said, and the system would reserve one entire lane on Interstate 95 to a single car until it reached its destination.
In contrast, packet-switched networks, which had conceptual roots in both Europe and North America, create a multi-user line. They break up each message into packets, stamp them with the message's address, and intersperse them with packets of other messages. On arriving at their destination, the packets are regrouped into the original messages.
In July 1968, Larry Roberts floated ARPA's request for the design of a 19-node packet-switched network, plus estimates of the cost of building a four-node network. Kahn put pen to paper, and with input from Severo Ornstein and several other BBN colleagues, devised the technical portion of BBN Technical Report 1763. Among other things, the proposal addressed how to control errors, sequence messages, and manage the buffer space that held incoming, unassembled messages. "Lots of this was not reducible to equations," Kahn recalled.
In late December, on his 30th birthday, Kahn got his best present ever. BBN has been chosen to build the world's first packet-switched network. Roberts had reviewed a stack of proposals more than 2 meters high and determined BBN had the best technical plan. He also liked their lean crew--about a half dozen for the design and building of hardware and software.
BBN project manager Frank Heart knew that even in Cambridge "capabilities of people often varied by factors of 10 or more." He chose small, elite teams, and persuaded Kahn to stay past his two-year leave from MIT and mastermind the setup.
Sleep apparently was optional. The first node (software, some hardware, and interfaces) had to be designed, built, and installed at the University of California at Los Angeles (UCLA) on Sept. 1, 1969, nine months away.
(c) Copyright 1996, The Institute of Electrical and Electronics Engineers, Inc.