Marc Slemko <marcs@???> probably said:
> The sender_verify_fixup option is horribly broken and should _NOT_
> exist or, at the very least, should have a big warning about it in the
> documentation so people aren't tricked into using it without knowing the
> nasty consequences and how badly it violates accepted practice.
The documantation says what it does.
> It is explicitly designed to take addresses from the headers and stick
> them in the envelope sender. That is a big no-no, since it causes mail
> loops in cases where the envelope sender was explicitly set to avoid them.
*sigh*
When people abuse the envelope sender and try to make it "invalid"
deliberately, things break. What do you expect ?
> This is especially true with qmail, since it uses an invalid (on purpose)
> envelope sender of "#@[]" for what it calls "double bounces". This
> invalid address is designed to ensure there is no way the message can
> bounce back and cause another bounce, etc. and create a loop. You can
> argue all you want about it using an invalid envelope sender, but the fact
Yeah, it's broken. This is what the null envelope sender is for, and
specified for in the RFCs. If you want to whine, go whine to the
authors of the MTAs that bounce messages with null envelope senders.
> is that the sender_verify_fixup option is what is broken in a loop-causing
> way by taking information from the headers of the message. It is not
> appropriate for a mailer to be trying to make such guesses in areas that
> have such critical consequences if you guess wrong.
It's not appropriate for a mailer to be producing deliberately wrong
envelope senders.
sender_verify_fixup is a well documented and very useful in some cases.
P.
--
pir pir@??? pir@???
