[ih] early networking

[John Day]

Yes, last summer I was reading Ian Toll’s Volume 3 of the War in the Pacific and it was mentioning that how the long delays for relaying radio messages through Pearl affected the outcome of battles. Odd, I thought. Then realized yes, they were torn-tape systems, at least at the relays. And message switching just automated that. I remembered from the early days of the ARPANET how the military was adamant about priority and precedence. And stories of important messages from an admiral being delayed behind long messages over 70 baud channels. 

Then the light dawned, as the video relates, message switching was analogous to FCFS batch processesing. Packet switching was analogous to multiprogramming (timeslicing) round-robin scheduling. (To continue the operating system analogy, long messages take a little longer but the completion time for short messages is shorter.) And virtual circuit was round-robin with contiguous memory allocation, and datagrams were a tool for exploring the next step,  but because they handled the immediate problem that step was never taken.

Recently, in exploring the papers of Derek Barber and Donald Davies, we have come across papers where Davies states explicitly that his inspiration for packet switching came after attending a 1965 IFIP Conference and hearing all about timesharing systems. He came back and told Derek that that was what they should do for communications. What is curious is that in the better known accounts by Davies on the history of packet switching he doesn’t mention this.

I have always believed that networking was much closer related to operating systems than to telecom. It is primarily a resource allocation problem. Telecom is really only a concern of the physical media, and the data comm protocols (link layer) turn out to be a degenerate case of the layers above. 

It is also interesting that reading Baran’s report that there is really nothing about this. He does cover issues like routing, congestion, etc. but it is more to check off the boxes that what he is proposing is feasible. His focus is distinctly on survivability, and the military uses of a such a network. Davies, on the other hand, was not involved with military research at all, their focus was more on these resource allocation issues (although I doubt he would have used that characteristics at the time). I have also been told by members of the CYCLADES project that their focus was on dynamic resource allocation (which dovetails nicely with Davies focus) and was one of the things they couldn’t get France Telecom to see. Because CYCLADES was shut down so early, they never got the chance to explore what they had seen.

Also remember that Peter Denning had shown in 1968 that static allocation of buffers required orders of magnitude more memory than dynamic allocation. We had seen this in the first OS we wrote where we used dynamic allocation, not because we were so smart but because we didn’t have enough memory to do anything else. ;-) It worked beyond our wildest imagination.

Take care,
John

> On Apr 8, 2024, at 18:05, Vint Cerf via Internet-history <internet-history at elists.isoc.org> wrote:
> 
> interesting pre-Arpanet/Internet history
> 
> https://www.youtube.com/watch?v=XFkwWZ6ujy0
> 
> -- 
> Please send any postal/overnight deliveries to:
> Vint Cerf
> Google, LLC
> 1900 Reston Metro Plaza, 16th Floor
> Reston, VA 20190
> +1 (571) 213 1346
> 
> 
> until further notice
> -- 
> Internet-history mailing list
> Internet-history at elists.isoc.org
> https://elists.isoc.org/mailman/listinfo/internet-history