Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site mit-eddie.UUCP
Path: utzoo!watmath!clyde!burl!ulysses!mhuxr!mhuxt!houxm!whuxl!whuxlm!akgua!gatech!seismo!harvard!think!mit-eddie!gds
From: gds@mit-eddie.UUCP (Greg Skinner)
Newsgroups: net.mail
Subject: Re: Mail problem, ihnp4 -> tektronix
Message-ID: <801@mit-eddie.UUCP>
Date: Thu, 19-Dec-85 13:52:46 EST
Article-I.D.: mit-eddi.801
Posted: Thu Dec 19 13:52:46 1985
Date-Received: Sat, 21-Dec-85 06:41:44 EST
References: <17623@styx.UUCP> <3080@sun.uucp>
Distribution: net
Organization: MIT Lusers and Hosers Inc., Cambridge, Ma.
Lines: 43

Perhaps Gary Murakami or Peter Honeyman would like to expand on this,
but I think I can shed some light on the problem.

The Sys V version of pathalias generates a long table of paths with
their associated costs.

> ihnp4 uses pathalias (or a variant of it) to do routing. Unfortunately, they
> seem to be getting more and more confused over time, with innnacurate or
> inefficient routes being generated. Every so often, ihnp4 used to
> believe that the fastest route to nsc was ihnp4!cbosgd!nsc even though
> nsc talked directly to them. Now, from my mail, it seems to think that
> ihnp4!cbosgd!sun is the fastest route to sun, even though sun talks to
> ihnp4 ... sigh.

As I recall, the cost of ihnp4<->cbosgd is cheaper than ihnp4<->nsc
since the phone call (if it even is a phone call anymore -- it might be
datakit) is cheaper (cornet) even though the actual rate might be
DEMAND.  Therefore, cbosgd appears first in the table.  The map data at
ihnp4 may reflect a very low cost from cbosgd<->nsc which may be even
less than that of ihnp4<->nsc so the ihnp4<->nsc entry never gets in the
table.

I remember running into a similar problem when I brought pathalias up on
houxm and I was trying to force mail to go to the ARPA Internet through
uucp links to an ARPAnet site.  Pathalias insisted that I take some long
route through cbosgd and ucbvax to get there.  What I did was after
pathalias generated the table I set the value of my ARPAnet gateway to
less than that of cbosgd, so that it would always be favored.  Probably
to get the desired paths you want to have you will have to modify the
UUCP maps to reflect that your own site has a cheaper direct route than
DEMAND.  Unfortunately this has the bad side-effect of favoring harvard
for uucp sites which it is connected to which I might have cheaper
routes to.

Who's in charge of ihnp4 now?  A quickie solution is to delete the table
entry for cbosgd!nsc and set the cost of nsc to that of any other DEMAND
site, which is what you want.
-- 
It's like a jungle sometimes, it makes me wonder how I keep from goin' under.

Greg Skinner (gregbo)
{decvax!genrad, allegra, ihnp4}!mit-eddie!gds
gds@mit-eddie.mit.edu