🚀 go-pugleaf

In article <1992Aug9.125252.3335@eua.ericsson.se> per@erix.ericsson.se (Per Hedeland) writes:
>I thought it would be simple, having seen (but never used:-) an attempt
>at implementation in some old sendmail.cf. E.g., assuming only a normal
>"stupid" sendmail, define a class L containing all locally known names
>(aliases + userids), and then replace the normal selection of the local
>mailer with:
>
>R$=L			$#local $:$1			known local names
>R$+			$#ether $@$N $:$1<@$N>		forward unknown names
>
>where $N is the name of the host to forward to.
>
>This works as a first approximation, but falls flat on its face in the
>presence of e.g. a not uncommon .forward file:
>
>\user
>"|/usr/ucb/vacation user"

  Train your users to create their '.forward' as:

	user
	|"/usr/ucb/vacation user"

  The training shouldn't be too hard.  Your users are motivated by the
need to not have their mail bounce :-( .  I think most sendmails
allow forwarding to oneself (unquoted) provided that the self address
is on the first line of '.forward'.

  For pipes, only the '|' need be outside the quotes.  This is enough
for sendmail.cf rewrite rules to recognize and send to the local mailer.

>I'm beginning to think this is really impossible with a "stupid"
>sendmail, and perhaps even with IDA (looking through some of the
>comments in the IDA Sendmail.mc regarding MAILSERVER - which is *not*
>what I want to do, but it has similar problems) - but I would be happy
>to be shown otherwise...

 Speaking of IDA, check out mp.cs.niu.edu:/pub/cf/simple.cf  .  It is
designed to forward most mail to a server.  It is simple minded, and
assumes it only needs to deal with properly formatted addresses (since
everything else is forwarded to the server).  If the server sends back
mail for a local user, it will be delivered locally, even after '.forward'
expansion, this based on the assumption that mailing loops are a bad
idea.  It does require an IDA sendmail for some of its features
(most particularly the mailing loop avoidance strategy).