Architects of the Net of Nets - Part 2
by Dick Reiman, Historian
Bob Kahn and Vint Cerf were working independently on the Pentagon's ARPA research project to develop a "packet switching" network, which would decentralize communications. Vint Cerf worked to develop a protocol so that four different sites or nodes of a network, each with a different computer host, could be interconnected or netted together.
Meanwhile, Bob Kahn was working feverishly to meet the deadline of September 1st, 1969, for the delivery of minicomputers called Interface Message Processors (IMPS). These were to be ready to be connected to dedicated high-speed telephone lines. BBN Corporation of Cambridge, Massachusetts, Bob Kahn's employer, stuck to the schedule and shipped the first IMP to California before the Labor Day deadline. The IMP weighed 450 kg, was the size of a refrigerator, and consumed 2100 watts. Its design was based on a Honeywell 516 minicomputer. It had a 100-us clock and could process 850kb/s. This then became panic time for Vint Cerf's group at UCLA as the developing of a protocol was lagging.
The IMP's had their origin in 1967 at the University of Michigan when ARPA's Larry Roberts decided that burst of data or "packets" would be sent by an IMP over the telephone line, and received by an identical IMP at the receiving end. The IMP would then interface with the host computer at each node of the network. Packet switching programs could be written for a single type of IMP computer, while the host computer, each of different manufactures, would view the IMP as a "black box". The concept of two (later more) steps at each node was called "layering". Host computers varied between a Sigma 7 with SEX operating system, a Scientific Data Systems 940 using Genie, an IBM 360/75 using os/mvt and a Digital EQUIPMENT CORP. PDP-10 with Tenex operating system.
Graduate students at UCLA led by Stephen D. Crocker devised the host interfaces. Their Network Control Protocol would connect host to IMP subnetwork and also to each other. The ARPAnet program was, however, both inflexible and didn't extend to future networks, and BBN was reluctant to share its software with the UCLA group. Enter Cerf, a key designer at UCLA, and Kahn of BBN to work things out.
"Vint and I hit it off professionally", Bob Kahn recalls, "We both liked really bad puns". Kahn traveled to California to assess the system. Said Cerf: "He would ask for software to do something. I would program it over night, and we would do the tests." Cerf continued: "we would crash the network trying to stress it. It exhibited behavior that Bob Kahn had expected, but other didn't think could happen."
The Arpanet began with 50-kb/s links across copper wire telephone lines. IMPS broke the messages from the hosts into 1000 bits, and to each packet added a standard header comprising the destination address, the source address and error checking. Then the packets were sent out. This plan met the objectives of no central routing control and could adapt to down lines. The Internet, however, was still several developments down the path to success.