Marc Haber wrote:
(CCing exim-dev for follow-ups)
> While Envelope-Sender is defined to be added by the last mail server
> in the transmission chain, it is probably a good thing to have that
> information available for servers earlier in the chain. That's why I
> think that it only makes sense if it is widely deployed.
I agree that it's really nice for tracing/debugging.
As for the spamd interface, there is something else: Only a single
recipient will be listed in the Received: header. It's easy to change
that, of course, but I think we don't want to end up having dozens of
recipients there. Not to mention the privacy issues, as some of them may
be bcc-recipients.
Looking at the code in spam.c and the spamd interface, it seems easy to
solve this: Send the recipients in a Envelope-To (or something else,
spamd is quite tolerant there) to spamd right before the mbox_file, so
spamd recognizes it as a header belonging to the message. (The same
could've been done for the envelope-sender, but it's not of much
relevance now.)
Other env-to headers should probably be stripped before, but I'm not
sure if that's really needed (will SA stop after the first found env-to
header?). The added complexity (if there's not already code for it in
Exim) is not worth it, I think. At worst, somebody would put a
whitelisted address in there, but if these are known, you are lost anyway.
Any concerns/comments about that?
