[Mailman-Developers] Mailman 2.0 usage notes...

Les Niles les@2pi.org
Wed, 11 Apr 2001 12:13:32 -0700

On Wed, 11 Apr 2001 10:42:43 -0400 barry@digicool.com (Barry A. Warsaw) wrote:
>    cvr> You can't assume this to be true for third level
>    cvr> domains. Overseas, it's NOT true -- because in many cases,
>    cvr> the third level of the domain is the equivalent to the second
>    cvr> level in .com. So if you assume that fred@anglican.co.uk is
>    cvr> the same as fred@catholic.co.uk, you just messed up badly,
>    cvr> because anglican and catholic are different companies. You
>    cvr> run into a rathole here REAL fast, because the assumptions
>    cvr> that are true for .com, .edu, and .net aren't true for the
>    cvr> country domains, where in many cases, it all bumps out one
>    cvr> sub-level (foo.com.br, boo.edu.pl, etc)
>Again, good point.  Actually SMART_ADDRESS_MATCH would only
>equivalence fred@catholic.co.uk with fred@co.uk, but that's not right
>either.  I believe all this stuff was added way long ago to help out
>misconfigured mailers, so maybe that's not as important these days.

How about a goodness-of-match algorithm?  fred@catholic.co.uk
matches itself with an error of 0, matches fred@co.uk with an error
of 1, matches fred@anglican.co.uk with an error of 2, and matches
everything else with an infinite error.  If the address in the
bounce has a unique minimum-error match with a subscriber address
(there's only one address that matches with that minimum error)
then that is used, otherwise it's considered unmatchable.  It seems
like this would handle most of the common cases and would very
rarely screw up, since bounces ought to have been generated by some
subscriber.  (OK, it would be nice to also recognize that
fredsmith@foo.bar and fsmith@mail.foo.bar are the same person, but
that's getting a little carried away....)