I have running been the latest Spamassassin via spamass-milter on my
FreeBSD 6.2 system - latest ports-current versions. I recently added
milter-greylist and kludged the port to build version 3.1.4 from 22 Jan
2007 (someone needs to chmod 755 configure). I also use SARE rules and
FuzzyOCR with SA. This works pretty well on my small server receiving
around 1000 e-mail's a day. I have noticed, however, that the
milter-greylist's auto whitelist is most spam. I have spamass-milter
set to 550 reject spam at score 15 and above. It would great to purge
whitelist entries for such obvious spam. I am quite willing to switch
to another Spamassassin milter if necessary.

A better question is can two milters communicate with each other easily
while processing the same message?

In case I describe in this thread Spamassassin's milter runs first
(milter execution order can be easily set in sendmail.cf) and during
the eom callback, the greylist milter wants to know if the Spamassassin
milter has bounced the message with a 550 code.. The Milter API does
not seem to have a way to find this rather basic fact out.

toonerh wrote:
> In case I describe in this thread Spamassassin's milter runs first
> (milter execution order can be easily set in sendmail.cf) and during
> the eom callback, the greylist milter wants to know if the Spamassassin
> milter has bounced the message with a 550 code.. The Milter API does
> not seem to have a way to find this rather basic fact out.

I'm not a programmer, much less a milter programmer. However, I can
tell you that I often see in logs where a milter that executes earlier
in the list rejects a message causing a subsequent milter to get a bit
annoyed at the fact that it has no data to filter. This would lead me
to assume that subsequent milters can deduce if a message has been
rejected by the fact that they do not have any data at a particular
point in their processing.

I may be *WAY* off base with this so, verify or refute this on your own.
However this may give you a direction to look.



Grant. . . .

On Jan 24, 5:41 pm, "toonerh" <t...@hedges.com> wrote:
> A better question is can two milters communicate with each other easily
> while processing the same message?
>
> In case I describe in this thread Spamassassin's milter runs first
> (milter execution order can be easily set in sendmail.cf) and during
> the eom callback, the greylist milter wants to know if the Spamassassin
> milter has bounced the message with a 550 code.. The Milter API does
> not seem to have a way to find this rather basic fact out.

Sendmail stops all milter processing for the message as soon as the
milter rejects the message. All other milters are sent the abort
command.

At which point does the greylisting milter perform greylisting? Which
callback is this?

On Jan 24, 5:41 pm, "toonerh" <t...@hedges.com> wrote:
> A better question is can two milters communicate with each other easily
> while processing the same message?

Not with sendmail's libmilter. Of course, you can code your own methods
in.

However, with milter-rrres patch you could have one milter set macro
values that are read by other milters.