So just thinking about this, EX_NOUSER is a 5xx code, which is a permanent failure. So the MTA would immediately bounce the message back. As it happens, someone would have received the message, and someone else would not have. But the person who *doesn't exist at all* would not be stuck in the queue -- and actually, *that* is the silent droppage -- it usually takes 4 days for a persistent temporary failure to bounce and often an administrator will slash and burn piles of stuck messages if they begin to gum up the delivery queue.
So EX_NOUSER or something similar seems more appropriate... Note also that even LMTP is susceptible to this problem if an address resolves to a list of other addresses, and one of them doesn't exist. Aaron Ilja Booij <[EMAIL PROTECTED]> said: > > Aaron Stone wrote: > > > I took an educated guess at this, figuring that it would be better to have some recipients get the > > same mail more than once than it is to have a message dropped for others. Multiple receive is > > annoying, but silently dropping is unacceptable. > > > > If EX_NOUSER is a better response, go for it! > > I guess you have a point when returning EX_TEMPFAIL then :) > > > > > Best of all might be finishing up the code for returning a human-readable response on stdout. > > This way, we can give EX_NOUSER and also give a message alerting the sender that it *may* > > have been received or *may not* have been received, and the sender should either send again or > > confirm the message by some other means. > > This still sounds somewhat dirty. I'm really wondering what the correct > behaviour would be. If it were a black box, I would expect to get only a > failure message for the failed recipient, but I wouldn't know how to get > to that behaviour without resorting to using LMTP. > > Ilja > > > > _______________________________________________ > Dbmail-dev mailing list > Dbmail-dev@dbmail.org > http://twister.fastxs.net/mailman/listinfo/dbmail-dev > --
