🚀 go-pugleaf

Peter Honeyman:
>>i've seen sendmail hacks (from waterloo?) that let rewrite rules munge
>>addresses with $>program.

Kyle Jones:
>I was told of this several months ago and thought it was a great idea.
>I was about to implement it here until I realized the amount of
>overhead it would entail per message.  Sendmail calls those rewrite
>rules a LOT.

Me:
I assume that you wouldn't be using such a trick in ruleset 3, only
after you'd munched and crunched the address down to size and were
left with, say, foo!bar and no direct link to the machine foo.  Under
the circumstances you either route or reject the mail, and if you use
a program to find routes it doesn't seem to me make much difference
whether sendmail exec's it or someone else.

I tried the $>program hack and didn't like it for some reason---I
think it's because there's little space left for magic cookies like
that in sendmail (or I was tired when I prototyped it and the code was
lousy).  I now use the $[ and $] canonicalization operators and allow
alternate ``name servers'' (either a program, or a dbm file, or an
actual network service) to be specified per ruleset.  If you have to
talk to really bizarre mail systems this is a big win---you can
declare `tr Rr Ww' to be a nameserver for .BARBARA-WALTERS.COM, for
example.

Joe Chapman
harvard!ima!joe