1965: The promise of packet switching
July 6, 1999
by Mary Brandel
(IDG) -- The time: The mid-1960s. The place: The National Physical Laboratory (NPL) in England. Donald Davies, superintendent of the computing sciences division at the NPL, was studying the concept of time-sharing. He began to realize the inadequacies of the data communications capabilities of the day. Unlike phone calls, where you needed an established circuit with a fixed bandwidth for the duration of the conversation, computer communications were more bursty, with long periods of inactivity.
Meanwhile, Rand Corp. in Santa Monica, Calif., had been asked several years earlier by the Advanced Research Projects Agency (ARPA) to develop a communications network that could survive a nuclear attack. "It was the height of the Cold War, and any attack by the U.S.S.R. would also take out the total U.S. telephone system by its collateral damage -- and we would be without military communication," says Paul Baran, who was instrumental in developing the ideas for the network at Rand. The phone system was so highly centralized that if even a small portion of the main plant were physically damaged, all long-distance communication would be blocked.
Two men, two very different projects. But Baran and Davies reached strikingly similar conclusions. Independently, both developed the concept of what we now know as "packet switching," a term Davies coined in 1965. Baran's term was "distributed adaptive message block switching." As he put it, "Davies chose a wonderful name."
Packet switching breaks big chunks of data into smaller units, called packets. The packets are transmitted individually using the fastest route available on the network. Each packet holds information about its origin and destination. Once all the packets arrive at the destination, they're recompiled into the original message.
With regular telephone service, which is based on circuit switching, transmissions are sent serially over a dedicated line. By comparison, packet-switched networks are more resistant to failure because only problem packets need to be retransmitted, not entire messages. Plus, if packets run into downed computers or cut lines, they find another path.
Packet switching is also cheaper because you can build networks out of less reliable parts because the overall system is far more reliable than any of its pieces, Baran says. No wonder most network protocols today, including TCP/IP, X.25, and frame relay, are based on packet switching.
Despite the similarities, the work of Davies and Baran took very different routes. Baran described his ideas in a landmark 1962 report, "On Distributed Communications," and then spent the next few years defending it. AT&T Corp., for one, was opposed to the idea of packet switching -- particularly because it required a digital network, Baran says. AT&T, which had a communications monopoly at the time, didn't welcome the competition brought by a digital provider. Plus, the vendor had only experimented with digital technology and didn't have much experience with it, he explained.
"The initial response was, 'The kid's crazy. It couldn't possibly work,' " Baran says. "At one of the early meetings with AT&T, one of the older guys said, 'Wait a minute, son. Did you say you opened the switch before the transmission got all the way across the country?' And he started explaining to me how the telephone worked. It's pretty hard for a guy who spent his whole life with analog to comprehend or take [digital] seriously."
The big break came in 1966. That was when two of ARPA's program directors, Robert Taylor and Lawrence Roberts, decided to build a network connecting all of the agency's universities and research-and-development firms. "They could have done it with centralized time-sharing, but they chose packet switching," Baran says. The result -- the Arpanet -- laid the foundation for today's Internet.
Meanwhile, back in England, Davies in 1965 dreamed of creating a national packet-switched network that would provide low-cost data communication across the U.K. However, with much less funding than the $20 million Arpanet project, he had to limit his work to a prototype network with a single switching node at the NPL. That network was completed in 1969 and lasted until 1986. Davies also gave lectures on packet switching, including one in Gatlinburg, Tenn., where ARPA's Roberts first became excited about the idea.
Did it take two great minds to make packet switching the most prominent network transport method available? Or was it inevitable that someone, somewhere would arrive at the concept? Baran isn't surprised that two minds thought alike. "There are only so many ways of doing this," he says.
Global effort to push Net protocol
RELATED IDG.net STORIES:
N+I: Cisco hails packet-based voice, data Internet
National Physical Laboratory (NPL)
|Back to the top
© 2001 Cable News Network. All Rights Reserved.
Terms under which this service is provided to you.
Read our privacy guidelines.