'The Web was invented by the British but exported to the US... We don't want that to happen again.'

Nathan Newman nathan.newman at yale.edu
Mon May 22 16:18:52 PDT 2000


On Tue, 23 May 2000, Rob Schaap wrote:


> So, if the ARPAnet was not the foetal internet, and if that was not a US
> DoD project that first saw the light of day in 1969, well, what's the real
> story?

It went back and forth. I have a quick summary of the story at:

http://socrates.berkeley.edu/~newman/chap2.html#b

Here's an excerpt for the heart of the back and forth and a few of the major players involved: ARPA, RAND, AT&T, the military, and Britain:

"As head of ARPA's IPTO division, Taylor was frustrated that even as he had the power of several different computer systems in his office , all had different computer terminals and could not communicate. Taylor decided that funding a computer networking project should be a high priority, especially since researcher around the country often needed to share expensive computer resources at locations far from their primary research location.

At the same time in the early sixties, researcher Paul Baran had begun planning how to build the technology necessary for the goal of networking computers. Baran worked at Rand Corporation, a company setup to monitor and preserve the US's operations research capability, where he worried about the survivability of US communication networks in the case of nuclear war. Modeling his ideas partially on the redundancy of neural networks in the brain, Baran envisioned the movement from analog signals to digital signals that could perform in such a networked system of digital transmission. Instead of a central switching node where a wire between two points would be reserved specifically for sound signals for that conversation, such a system would be a "distributed network" with each node connected to its nearest neighbors in a string of connections, much like the child's game of telephone. More dramatically, messages would be broken down into parts, travel the network and be reassembled at the opposite end in a system called packet switching. This would allow fuller use of all lines in the network instead of holding lines open from end-to-end for each message. Each node would keep track of the fastest route to each destination on the network (an d be constantly updated with information from adjoining nodes) and help route information without need of central direction.

RAND was enthusiastic about Baran's ideas but when AT&T was approached about its feasibility, AT&T executives dismissed the idea and even refused to share information on their long distance circuit maps; Baran had to purloin a copy to evaluate the ide as which he and RAND were convinced were right. Based on RAND's recommendation, the Air Force directly asked AT&T to build such a network but AT&T still refused saying it wouldn't work (although a faction of scientists at Bell Labs did support the idea.) This may have been technical myopia by the business-oriented executives but it was an economically self-interested myopia derived at least partly by a political box in which the federal government itself had put AT&T. Such a distributed network threatened (and today does threaten) the central economic assets of the telephone industry: central computers and central switches. On one level, AT&T's resistance highlights the fact that corporate research labs, the main alternative to long-term government funding of technology, rarely if ever invest in fundamental technology that will likely undermine the economic monopolies they currently enjoy. Compounding the problem for AT&T was the fact that the first winds of deregulatory attack on the company were b lowing and, specifically, the company was increasingly barred from selling anything having to do with computers attached to the phone network. As will be detailed in Chapter 5, this emerging deregulation and its political division between the phone network and the increasing computerization of telecommunications would have perverse consequences for both technology and regional economies.

The Air Force contemplated building the network by itself but it bogged down in internal organizational problems in getting the project off the ground. In the meantime, British physicist Donald Davies had begun promoting a similar idea of a computer network with "packets" of information. He soon learned of Baran's similar ideas and was encouraged enough to get support by the British Post Office, which ran the telephone system in Britain, to support the concept. Without the political issues of deregulation, the state-run phone system in Britain just treated the project as a simple demonstration project. In 1968, the first computer distributed network was established on computers all located at the National Physical Laboratory where Davies worked.

Taking off on Davies' example, ARPA began its own networking project. Larry Roberts from Lincoln Labs was hired to oversee the computer networking project. There was a certain hostility from many East Coast universities to sharing scarce computing resources which led to the network starting out at four west coast sites: University of California Los Angeles (UCLA), UC Santa Barbara, University of Utah, and the Stanford Research Institute's Augmentation Research Center. The idea was to install a new computer at each site as part of the network: this would avoid direct incompatibilities in the network and allow each campus to focus on a separate interface between their regular campus computers and the local network computer.

In 1968, ARPA advertised a bid for building the subnet computers (which would be called "Interface Message Processors" or IMPs). IBM and other big computer companies declined even to make a bid, saying it was not possible at a reasonable price. Like AT&T, this was partially the myopia of those grounded in older technology but it was also a self-interested economic fear of the new minicomputer technology, supported by the federal government, that was challenging the dominance of companies like IBM. IBM and others rightly feared that networking would make many government agencies and businesses rethink the need to actually own their own mainframe computer.

While the defense contractor Raytheon almost won the contract, in the end Licklider's old consulting firm Bolt Beranek and Newman (BBN) convinced ARPA that it's relatively small operation (600 employees) could do the best job. With its ties to MIT, including a workforce made up largely of MIT graduate students, BBN had a good case for implementing technology largely developed at that university. With a large ($1 million) contract, it was able to take on a project it could never have done on its own w ithout the guaranteed market the government was providing.

By October 1969, the network connection between UCLA and Stanford was established and within months, all four "nodes" plus BBN itself were online. By the time the network was demonstrated publicly for the first time at the International Conference on Computer Communications in October 1972, there were twenty-nine nodes in the network (dubbed at this point ARPANET) clustered largely in four areas: Boston, Washington DC, Los Angeles and San Francisco. What would evolve into the Internet had been born."



More information about the lbo-talk mailing list