On 24Nov 2017, at 6:52 PM, Paul Ruizendaal <pnr@planet.nl> wrote:

Found a post by dmr about this in the 'net.unix-wizards' newsgroup:

It would seem that my understanding of it is indeed how dmr designed it.

At the end of the post there is an example: "For example, from my machine, I can type "rlogin purdy" and connect to a Sequent machine running 4.2; the TCP connection goes over Datakit to machine "research" where it is gatewayed to a local ethernet that purdy is connected to."

I'm not quite sure how on the Sequent machine the Datakit channel would be hooked up to the 4.2BSD TCP stack. Perhaps something equivalent to e.g. 'SLiRP' was used, perhaps there was a pseudo device that hooked into the network interface layer.


Nope, not IP over Datakit, as I recall.  It was quite interesting to work at a site (Bell Labs) where there were two distinct network technologies.  

We connected to a lot of familiar services through Datakit using the application level.  For example, the smtp client in upas connected to an SMTP service thus:
       if ((s = ipcopen(path, "")) < 0) {
where path might be one of:

On V10 Unix and in Plan 9, these were handled by stuff that Dave Presotto wrote.  I distributed this code throughout internal AT&T.  

The first is the obvious tcp connection.  Datakit connected to a server running in V10’s equivalent to inetd.  The last one was a proxy connection, a circuit-level gateway running on a particular host that I supported for many years.  It had a little protocol and initiated the tcp link from the proxy machine.

These proxy machines had some interesting features.  For one, I could log the destination of volume of each connection.  For another, all connections to a sensitive site (think porn) seemed to come from the proxy host.  Some porn sites showed their top ten users, and several major corporations were represented there.

This library was socks about seven years before socks, originally written by Presotto and Howard Trickey.  The relay program was originally called “gated”, but that wouldn’t do after a while.  I renamed it “proxyd”, and that is the first use of “proxy" in this context that I am aware of.

 If you were on AT&T’s intranet and wanted to connect externally, you ripped out the entire socket dance and replaced it with an ipcopen call.  I also distributed common modified clients, like ptelnet, pftp, pfinger, etc.  

I still have all this code, and I suppose it ought to go in an archival repository.  I can’t imagine that AT&T/Lucent/Alcatel/Nokia would care at this point. Anyone want it?