Garry,
Interesting idea!
Though the IMPs only need 2 UDP ports per 'leased line' connection. So if someone runs his own IMP (with his host machines attached to them locally) that'll work differently.
A host computer connects to his IMP with two UDP ports. Following your idea, the IMPS would all run on the central server and a user's machine (say, a PiDP-1 :-) ) would get assigned to a host port to an IMP on the server.
So two different models to connect up. Both, admittedly, better than what I have on my test server. (
https://obsolescence.dev/arpanet_home.html) Which is one connection open for an external IMP and, I believe, 8 for connections from a host computer to an IMP already running on the server.
It is all a matter of config files. For now, I focus on two 'models': one public test Arpanet on that server, very much not set up in the best way. And one model, cloning this on a local system so someone has the Arpanet-in-a-box running locally just for himself. Aaah, the whole Arpanet all for yourself :-) No people.
The fun, of course, is in using that Arpanet to have your MIT and Stanford machines accessible to play with.
Kind regards,
Oscar.