gds@mit-eddie.UUCP (Greg Skinner) (01/11/86)
I have watched this discussion with some amusement, because a long time
ago I remember a discussion in net.mail where this very topic was
discussed, except pathalias hadn't been written yet. We were talking
about what might happen if smart mail routers attempted to rewrite
addresses, or if routing data which was out of date was being used.
This discussion also took place once after a version of the uucp maps
had been posted where a major site's entries were off, which caused some
strange paths to be generated.
Anyway, Fred Avolio is right, any mail router should not attempt to "fix
up" the path by rewriting it, inserting sites along the way which it
feels are "cheaper" or more "well-connected". (Robert Plamondon points
out also, that "well-connected" does not always mean you'll get your
mail through there all that quickly.) No offense to Gary & company, but
when I as at the Labs I alternated between sending my stuff through
ihnp4, allegra, cbosgd and a few others depending upon how my (houxm)
uucp and RJE queues looked for them. John Quarterman, I think, once
said that the biggest problem with UUCP is that there is a lack of an
adaptive routing service, which would do what Gary Murakami suggested:
form paths based on loading and queueing factors, and in-transit
hop-to-hop times.
At any rate, I would like to propse a way for incorrect path generation
to be suppressed. A new header, to be added to the mail message:
Routing:
which can have values which amount to arbitrary routing, loose routing, or
strict routing. If the arbitrary flag is on, routing can be decided by
any intermediary site in the path, including rewriting the address. If
the loose flag is on, an intermediate host should not attempt to rewrite
the path unless (1) the next host in the path is not connected, or (2)
there is a direct connection to a host further along in the path. If
strict routing is on, *no* attempt should be made by any intermediaries
to rewrite the path in any way. If any intermediary in the path is a
bad system name, the mail should be returned to the sender. Also, if
any given site would insist on rerouting, it should return the mail to
the sender informing the sender that the mail won't go through unless it
is rerouted. Then the user may re-send the mail through that site
(subject to any consequences of routing) or use another path which may
be better.
In general, if you suspect that your mail is not getting through certain
sites, it's best to use alternate routes. Unfortunately, it's hard to
determine what a "trusted" route is when you're a few hops away from
your destination.
--
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