Paul Johnson wrote:
> Accept it, only deliver to the negative user would be the sane way of
> handling it. Either that, or 550 it and say "Delivered to ..., not
> delivered to ..."
You're not understanding the problem. Think of it at the SMTP level,
we're filtering there, right? Ok.
rcpt to: foo@bar.com
rcpt to: baz@bar.com
data
blah blah blah....
.
FILTERING HERE.
Who's settings do you use? Even worse, which Bayesian database do you
use? You can use only one or the other. Once you filter that's it. So if
foo rejects, you reject for all. If foo accepts, you accept for all. There's
simply no two ways about it; that's how Exim's internals works. It has been
discussed at length, to death in fact, on Exim's list. If you're so inclined
a check of the archives there will give you more detail than you ever want to
know.
Regardless, SMTP time checks are system wide, not user specific. If users
want their own filters it will have to be after your system has accepted the
message. At that point it is no longer SMTP time checking and outside the
scope of both sa-exim and exiscan. At that point you're onto the MDA which is
either Exim's own internal routines controlled by a specially formatted
.forward file, procmail or what have you.
Personally my own spam prevention techniques filter things 3 times. Two
are SMTP time checks (clamav via exiscan, spamassassin via sa-exim) and
Thunderbird's Bayesian scanning. Obviously TBird's is well after smtp time. :)
--
Steve C. Lamb | But who decides what they dream?
PGP Key: 8B6E99C5 | And dream I do...
-------------------------------+---------------------------------------------
Attachment:
signature.asc
Description: OpenPGP digital signature