by Darius Kazemi, Jan 19 2019
This RFC is two suggestions from Kreznar on how to reduce network congestion, somewhat in the spirit of RFC-18.
His first suggestion seems to be to change how IMPs work so they consider the local state of the HOST machines when sending messages. So like if user A and B are logged into a HOST machine and the IMP has messages for them both, but user B is the one who is actively being addressed by the computer's memory, it would make sense to send the message for B first while user B's processes are in memory, then the message for A which the HOST machine could swap user A into its active memory. If the messages came in the reverse order there would be extra swapping, slowing down the computing experience for the users.
This seems like a really bad idea to me, since it would require the IMPs to have knowledge of the internal state of the HOST machines to some degree. That said, the internet has existed for literally my entire life and these folks were building it for the first time so... I'll let it slide.
The second suggestion is that maybe for transfer of large files between two HOSTs, the HOST could “lock” the program that it's running for the duration of the file transfer. Basically a way for both computers to say “we are going to prioritize this file transfer over anything else happening on the computer right now so we can quickly transfer the file between the two of us”. On a timeshare machine this would result in users who aren't involved in the file transfer having to wait for the transfer to complete before they can continue using the computer. That seems... also bad.
Both of these ideas aren't great. But again, no internet exists yet so I can forgive bad ideas! The whole point of RFCs at this early stage is to get as many ideas out there as quickly as possible. And it seems to be working.