Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP
Posting-Version: version B 2.10.1 6/24/83; site decvax.UUCP
Path: utzoo!linus!decvax!jmcg
From: jmcg@decvax.UUCP (Jim McGinness)
Newsgroups: net.mail
Subject: Re: Smart routing
Message-ID: <381@decvax.UUCP>
Date: Sun, 19-Feb-84 04:05:46 EST
Article-I.D.: decvax.381
Posted: Sun Feb 19 04:05:46 1984
Date-Received: Sun, 19-Feb-84 08:22:08 EST
References: <5524@decwrl.UUCP>
Organization: DEC UNIX Engineering Group
Lines: 28

As perpetrator of the path improver in use on decvax, sophomoric as it
may be, I feel compelled to say a few words in its defense.  Under the
circumstances, it's probably just as well that I cannot remember who
first suggested this method.

The path improver short circuits paths containing a uucp neighbor.  For
example, the path "siteA!siteB!siteC!user" is truncated to "siteC!user"
if decvax has talked to siteC recently.  Uucp (not just our version)
keeps a stats file containing, for each site, the time of the most
recent conversation or transfer.

The important qualities of this method are that it is local, automatic,
fairly safe, and uniform.  All of the information needed is available
on the machine doing the routing.  There is no database to be collected
and maintained.  Mail will not be mis-directed along paths that have
fallen inactive nor redirected into loops.  If everyone used it, it
would work even better.

It does have failings.  If it finds "vortex" in the path, then that
piece of mail will go to LA rather than to the E-net.  If someone sends
mail on a long loop to probe a path or provoke a poll, it will get
bounced at decvax rather than having the desired effect.  It doesn't
save us any money, just that of those sites we trim out of paths.

						Jim McGinness
	decvax!jmcg				Digital Equipment Corp.
	(603)844-5703				MKO2-1/H10
						Merrimack, NH, 03054