I know that LMTP defines a method for handling multiple recipients. I don't think that the pipe delivery interface does, and in fact I think it specifically is for one recipient only... unless we're parsing headers, in which case we might have a few recipients.
I'm inclined to say that any delivery error to an expansion is a delivery error for the original target address; at the very least, this ensures that everyone will get a fair attempt at the delivery. An annoyance might be multiple copies to those who recieved it in the first place and are receiving it again in the redelivery attempt hours / days later. Aaron "Christian G. Warden" <[EMAIL PROTECTED]> said: > On Wed, Feb 04, 2004 at 01:15:48PM -0000, Aaron Stone wrote: > > Based on my reading, I think that DBMail should not generate bounces at all. > > Rather, appropriate temporary-failure and permanent-failure return codes > > should come from dbmail-smtp and DSN codes from dbmail-lmtpd, along with > > additional descriptive text in both cases, and the MTA will do the rest. > > I agree that the MTA should handle undeliverable messages. The only > problem arises when dbmail-smtp is passed multiple recipients and only > some are undeliverable, or, if after alias expansion, some recipients are > undeliverable. > > > If this is the case, I can remove the bounce call from pipe.c right now; > > I'm working on refining the delivery chain to produce appropriate error > > codes for various failure conditions. Making sure to emit bounce codes > > for situations which would have called the bounce() function won't be > > too hard. > _______________________________________________ > Dbmail-dev mailing list > Dbmail-dev@dbmail.org > http://twister.fastxs.net/mailman/listinfo/dbmail-dev > --
