🚀 go-pugleaf

In article <V1L9VB3w165w@willard.UUCP> dawson@willard.UUCP (Willard Dawson) writes:
|Could the MX record on a given site still be made to point to the
|true best path to the UUCP-connected site?

MX records point to hosts, not to paths.

|If the MX record for willard.UUCP were to indicate a path of
|gatech.edu!vdbsan!willard!%s, or something similar, would that not still
|route mail in the exact same way that UUNET would route mail otherwise?

The MX record would have to point to "gatech.edu".  This is where the
biggest problem arises - how do you come up with the proper Internet
host at which to point the MX record?  Note that, at least in psuvax1's
pathalias data, the route to willard is "gatech!vdbsan!willard!%s" --
no mention of any Internet sites in that path at all.  So you need to
process the maps again, looking for sites that look like they may be
Internet hosts, and then make their MX'er the closest Internet host.
This means, at least, a new pathalias-like program.

Now, the next problem:  you have to make sure that all of these machines
are actually on the Internet.  How do you know that
gatech	=gatech.edu is on the Internet, while
wcfpc   =growler.redbeard.com is not?  You can do a DNS lookup, but
there are a *lot* of records to look up... this will probably end up
putting a significant load on your system (significantly more than just
pathalias).

Ok, are we done yet?  Nope -- how do we know that the Internet site we chose
is set up properly to accept mail to .UUCP?  Just because a node is on the
Internet and is in the .UUCP pseudo-domain doesn't mean that it can route
between the two properly.  You're likely to get horribly mangled From: and
To: addresses if you just decide to pick a random site as your gateway.


A new flag in the pathalias data, noting that one is:
a) connected to the Internet
b) willing to gateway to the .UUCP pseudo-domain (maybe a radius factor --
psuvax1 might be willing to gateway to hosts less than 2 hops away, or
something)
would solve the second and third problem.  Then the only problem left
would be writing a graph-traversal program similar to pathalias but to
find the closest Internet site to each UUCP host.  If you want to
select the closest Internet gateway to each site, this turns out to be
a problem N times as big as the pathalias problem, where N is the
number of Internet gateways registered.  The pathalias data from any
particular site will not necessarily give the closest Internet
gateway.  Let's say, for example, that willard connects to psuvax1.
UUNET's path would still be gatech!vdbsan!willard, but the closest
Internet gateway would actually be psuvax1.

And, of course, the new flag would be slow to propogate into the maps, and
would have the expected problems (not getting updated, etc.).

  Bill