On 4 Oct 2006, at 19:47, Andreas Metzler wrote:
Thanks, the patch seems to have worked for me, I have been running it
since the 29th of September.
Unfortunately the patch is wrong, I should have been more cautious
(not that it breaks postgrey however).
After looking deeper into the problem I've figured out this isn't a
postgrey bug.
I assume you, like me, were using postgrey with exim and blindly
followed the instructions contained in the README.
It turns out that postgrey requires the "instance" parameter, which
is given in postfix, but not in the supplied exim configuration example.
exim does not provide an identifier like this, but you can construct
one (see below).
How postfix can generate such an id properly is beyond me.
IMHO using "instance" is useless, as the message instance is
implicitly the "triplet" without subnet truncation.
This is the minimal required exim expansion:
set acl_m0 = request=smtpd_access_policy\n\
client_address=$sender_host_address\n\
client_name=$sender_host_name\n\
sender=$sender_address\n\
[EMAIL PROTECTED]
instance=$sender_host_address/$sender_address/
[EMAIL PROTECTED]
Can we expect a fixed documentation?
As a side note, I switched to postgrey from my own implementation
(http://maps2.ubiest.com/~wavexx/spamstats) in the hope for a cleaner
tool, _less_maintenance_, and possibly better performance, but I've
found postgrey not to be the best companion for exim (although it
works). A pity other greylisting daemons aren't maintained or use a
full db.
exim has a more advanced expansion mechanism that allows to construct
any kind of generic key you want instead of a "triplet" (making,
among others, --loopup-by-subnet useless for example). Much more
flexible for the advanced postmaster.
I was using this system in my implementation (decoupling the key),
for example, to build a helo/host key only, which turns out to be
more effective than the classical triplet as seen everywhere: less
delays for large systems, fewer entries, and proper rejection of
proxies.
But I guess you have to choose between having total control or to
focus on something else :)
--
To UNSUBSCRIBE, email to [EMAIL PROTECTED]
with a subject of "unsubscribe". Trouble? Contact [EMAIL PROTECTED]
