BBN - the contractor that built the first network - A research firm won the ARPANET contract, and the packet switch became a product because somebody had to manufacture it.

Bolt Beranek and Newman began in acoustics - the firm was hired to make auditoriums sound right - and ended up building the machine that made packet switching real. Winning the ARPANET Interface Message Processor contract in 1968 turned a funded idea into hardware somebody had to deliver, support and repair, which is a different problem from proving the idea works.

BBN did not set out to build networks. It began when Richard Bolt, running MIT's acoustical laboratory, won a 1946 commission to advise on the acoustics of the United Nations Headquarters, realised on seeing the drawings in 1948 that it was not a one-man job, and asked Leo Beranek to take it with him. Robert Newman joined in 1950, the name changed, and the firm incorporated in 1953 as an acoustics consultancy.

It entered computing in 1957 on a deliberately vague premise - that the company needed to be in that business to become known - and the vagueness turned out to be the strategy. J.C.R. Licklider joined that spring. The first machine arrived in 1958, DEC's first PDP-1 in 1960, and the arrival of the hardware is what let BBN hire John McCarthy and Marvin Minsky. By about 1960 the computer side was as substantial as the acoustics side, and the firm had built one of the first time-sharing systems.

In July 1968 ARPA put out a proposal for a device called the Interface Message Processor. BBN won it in January 1969 against much larger competitors, with a team of Frank Heart, Dave Walden, Bernie Cosell, Severo Ornstein, Robert Kahn and Will Crowther, and built four machines for close to a million dollars between September and December of that year. Those IMPs were the first generation of what the industry now calls routers.

The famous detail is worth stating precisely because it is usually told as a joke. The first message between the UCLA and Stanford Research Institute machines was 'LO' - the researcher was typing LOGIN and the receiving host crashed partway through. A network's first transmission was a partial word caused by a failure, which is a more accurate introduction to the field than any success would have been.

The reading that matters here is about institutional shape. BBN was a consultancy hired to build something, not a product company defending a market, and that is exactly why it could take the work: it had no installed base to protect and no line of business the new thing threatened. Almost every foundational piece of the early network came from an organisation with that property - a laboratory, a university, a contractor - and almost none from the companies then selling communications equipment.

BBN ↗

Sources