Chapter 6 - Networking: Arpanet 1969-1972
6.1 The Communications Subnet: BBN 1969
In January 1969, the BBN team began the painstaking tasks of fleshing out the design of the communications subnet. It had been agreed that the subnet would consist of minicomputer-based Interface Message Processors (IMPs) initially interconnected with leased telephone lines. Hosts would communicate with other Hosts by sending messages over the subnet. IMPs, in a process totally transparent to Hosts, routed a message by parsing it into up to eight packets with the destination IMP reassembling the packets into the message before delivering it to the intended Host. A message would consist of roughly 8000 bits, while a packet was limited to 1000 bits. Exactly how this was to work, reliably and error-free, was now the challenge. It was to be up to personnel at the Host sites to determine how sending messages rather than establishing circuits was to lead to a radically new means of communications between computers.
Given this architecture, a number of important issues remained to be specified; a routing mechanism, for example. The BBN response to the RFQ called somewhat vaguely for the “shortest-path routing, with certain metrics.” As Kahn recalls:
We had left the precise implementation open, because we knew it was an area we wanted to do more research. Paul Baran’s contributions to routing were very interesting, because they were among the first dynamic discussions of how you do routing. They were largely based on what he called “hot potato” routing. You have to understand where Paul was coming from. He was thinking of this kind of network as a survivable network from the point of view of military communications in the era of thermo-nuclear war. Herman Kahn had just written his book “On Thermo-nuclear War” which was getting everyone upset. So the military was looking for ways to deal with communications in very violent kinds of threat environments. I also think Paul was motivated almost entirely by voice considerations. If you look at what he wrote, he was talking about switches that were low-cost electronics. The idea of putting powerful computers in these locations hadn’t quite occurred to him as being cost effective. So the idea of computer switches was missing. The whole notion of protocols didn’t exist at that time. And the idea of computer-to-computer communications was really a secondary concern. What finally happened in networking was, in fact, very strongly derivative of what Baran was talking about, but he had not really laid a clear template for what it was.
Other issues that concerned Kahn, the principal architect, were congestion control, independence and deadlocks: How could the network be kept from being congested by too many packets? How do you build a system where each of the nodes would act independently of all the other nodes and still talk to each other without requiring global control? How to keep the network from coming to a grinding halt? The last concern, deadlocks, was more nebulous and became more “contentious.” Kahn believed this was an important issue, but was unable to convince anyone else.
Soon after winning the contract, BBN received notice the project specification had changed: each IMP had to accommodate the connection of up to four host computers, not just one. This change was significant and concerned BBN, for if the specification was subject to change, how would they ever complete the project on time? Roberts and Heart began meeting regularly. Roberts was concerned about the schedule as well, for if the host sites did not take it seriously, it would be difficult to motivate them to make the project the priority Roberts felt necessary.
Roberts and Frank Heart benefited from being from organizations that gave managers the “room to make it happen.” For when they made a decision, that was the decision. No other approval was needed. The high level of trust and integrity between the two organizations - grounded in the relationships among the many individuals who had been members of not only ARPA and BBN but also MIT and Lincoln Labs - facilitated project management. Roberts and Heart knew each other and knew they would likely be working with each other in the future. These cultural and organizational attributions became critical as Roberts and Heart pushed to complete the Subnet on schedule. Heart opines:
Larry at his level in ARPA, was very much in charge. Almost nobody was looking over his shoulder closely and, at BBN, I was a free agent. There was nobody looking over my shoulder either. So projects like this that are this hard, it’s very critical not to have too much back seat driving, and that’s one of the things about this project that was very unusual.