Re: UNIX listserver

[Matti Aarnio <mea @ utu . fi>]

>From: "F. Scott Ophof" <Ophof@CS.UWindsor.Ca>
> On Thu, 8 Jul 93 16:29:54 EDT Wes Morgan said:
> >What if I write an exact Unix clone of LISTSERV?  Arrgh.....
> 
> Which users & maintainers etc. couldn't distinguish from "Revised
> LISTSERV" and interfacing with the other "Revised LISTSERV"s exactly
> as they do with each other?  Of course you wouldn't use "Revised
> LISTSERV" code, or use "LISTSERV" as product name...  :-)
> I would very much applaud such a project, and point out that CREN
> currently is engaged in just such a project!  And I'd sincerely hope
> Eric wouldn't grumble at <listserv@..>-type addresses being used.

  No, he won't, however he does expect that such LISTSERVers do act
like Revised LISTSERVs, and one BIG snafu in it is  R-LISTSERV's
DISTRIBUTE-protocol, which is documented in RFC 1429. (Yes, it really
looks alike MVS JCL, so what ?)
Actually it is a sort of SMTP inside the mail body. R-LISTSERVs just
handle it better.

  The protocol is easy to implement, but how do you find the (near-)
optimal list of peers willing to run it ?  R-LISTSERVs do it by
centralized control, and monthly distributed routing database.
(They are run in the BITNET after all..)

  A month or two ago Eric made a contract with EARN to develope a
VMS version of the R-LISTSERV.   VMS because VMS Pascal is nearest
to the PASCAL-VS, that most of the R-LISTSERV is written today.


  I assume some of you are going to IETF at Amsterdam, could we
try to find a BOF for "Bulk-MXes" -- for a way to discover who
are willing to run bulk deliveries around the network.

Hmm.. Consider scenario:
	Host-A(Finland) a list of 1000 recipients of which 20 in Finland.
Now the usual thing is to start spiffing MX and A questions around
the network, and then deliver mail to each and all of them via individual
connections, unless by some miracle multiple recipients are at the same
host/mx, or then by getting into some arrangement with some sites abroad,
and define static routes to large areas of domain space via those..

	What I wish for is a simple scheme of finding willing
Bulk-delivery servers, much like R-LISTSERVs.  In scenario I usually
consider, next logical hop is in Sweden, which then splits to Central
Europe (say, Paris, Amsterdam, CERN, Bonn -- corners of EBONE), and
USA+Asia.

	Again, next step would go to - say USA (EDU that is):
How to locate regionals onto which to push the mail, or just use plain
ordinary MXes ?  If the scheme is multi-level, how does one avoid
unnecessary traversal to higher-levels ?

	How would such resource finding operate ?
Especially as it would be good to assume that people at networks know
their connection routes, and are thus better positioned to place such
routing information into the DNS.

> If I'm not mistaken, Tasos' intention was that "ListServer" be
> indistinguishable from "Revised LISTSERV" as to the common subset
> of functionality and user-interface.

  Well, let commands be the same, and their behaviour, but it can
still present a bit different outlook ;)

> Regards.
> $$\

	/Matti Aarnio	<mea@utu.fi>
			<mea@nic.nordu.net>