[ih] TCP adoption in 1984

[Larry Masinter]

>  I personally have no experience at all with PUP running over 10MB
Ethernet;

We been experiencing PUP over 10MB Ethernet using

    https://files.interlisp.org/medley/sources/PUP.pdf

 using the "Dodo" emulation of 10MB ethernet (
https://github.com/devhawala/dodo)
but a better encapsulation would be desirable.

I'm hoping to allow reenacting of early 80s networked applications like
Loops Truckin (network multiplayer game) and Colab (shared editing
environment).

One of the first problems we fixed (
https://github.com/Interlisp/medley/pull/327) was the 15-second delay
waiting for a response to the "What is my PUP hostnumber?" request to time
out when there was no ethernet enabled at all.

Larry
--
https://LarryMasinter.net  https://interlisp.org


On Tue, May 5, 2026 at 3:35 AM Noel Chiappa via Internet-history <
internet-history at elists.isoc.org> wrote:

>     > From: Bill Nowicki
>
>     > PUP (and I think ChaosNet?) did not require ARP
>
> I originally skimmed past this message, but on seeing it again, it has some
> confusions in it that I ought to correct, so there's no error in the
> record.
>
>
> First, I personally have no experience at all with PUP running over 10MB
> Ethernet; I'd never even heard of it being done. Obviously, if it were
> done,
> it would have needed something with the functionality of ARP, to turn PUP
> addresses into 48-bit addresses (for wrapping packets with, for
> transmission). I looked to see if there was such a thing, and found this:
>
>   When a PUP level one packet is transported by an Ethernet, or an NS level
>   one packet is sent on Experimental Ethernet, the level one host number
> cannot
>   be used as the level zero address, but rather some means must be provided
>   to determine the correct level zero address. Xerox solved this problem by
>   specifying another level-one protocol called {it translation} to allow
>   hosts on an Experimental Ethernet to announce their NS host numbers, or
>   hosts on an Ethernet to announce their PUP host numbers. Thus, both the
>   Ethernet and Experimental Ethernet Level Zero Protocols totally support
>   both families of higher level protocols.
>
>
> https://xeroxparcarchive.computerhistory.org/erinyes/lispmanual/EtherOverview.im!2
>
> So, something like ARP, but not ARP.
>
>
> Second, the CHAOS protocol (something different from Chaosnet hardware -
> the
> MIT AI Lab guys unwisely used the same name for both) definitely did use
> ARP,
> because ARP was created in part to allow the CHAOS protocol to run on 10MB
> Ethernets (same issue - where does one get the 48 bits from, given the
> 16-bit
> addresses in CHAOS protocol). DCP and I had CHAOS protocol _and_ IP in mind
> as clients for ARP when we designed it.
>
> DCP was a CHAOS guy; he was part of the MIT AI Lab crowd, and they wanted
> nothing to do with TCP/IP at the time when ARP was done - November, 1982.
> He
> wound up writing the RFC, not me, is all (I'm not positive why; I was
> probably too busy, or something like that). One can tell I had a lot to do
> with the design, because of the generality - an attempt to make sure it had
> the maximal lifetime - at which I think it succeeded.
>
>         Noel
> --
> Internet-history mailing list
> Internet-history at elists.isoc.org
> https://elists.isoc.org/mailman/listinfo/internet-history
> -
> Unsubscribe:
> https://app.smartsheet.com/b/form/9b6ef0621638436ab0a9b23cb0668b0b?The%20list%20to%20be%20unsubscribed%20from=Internet-history
>