[Spambayes] server side spambayes

Amedee Van Gasse amedee at amedee.be
Thu Nov 9 13:00:11 CET 2006


I am using spambayes for almost a year now, and I'm very happy with it.
Most spam gets caught (>95%), and no false positives after initial
training.

Currently I am using spambayes as a procmail filter, in ~/.procmailrc.
For training, I have a few imap folders: ham, unsure, spam, confirmed-ham,
confirmed spam. On average, 1 or 2 messages per day arrive in one of the
first 3 folders. I move them to one of the confirmed folders.
(This setup is described somewhere on the wiki)
A nightly cron job then trains on the confirmed folders.

This works great, but it's client-side filtering so it only works for me.
Recently my spouse showed interest in spambayes because the spam filtering
that our isp does is - honestly - crap. Too many false positives and too
many getting trough.
I'm also hosting a few (low traffic) mailing lists on my server, for
friends and for a few associations I'm in.

Because of this, I am interested in using spambayes at the server level.
I have found this page:
http://spambayes.sourceforge.net/server_side.html
with "postfix notes from Jonathan St-Andre"

I still have some questions.
If I understand it correctly, the only thing this setup does, is tagging
emails with a spam score, just like I am already doing in procmail. Is
that  a correct interpretation?
So it is still up to me to classify the emails in the correct ham/spam
folders using procmail?
Is it possible to block emails at the server level when they have a spam
score above a certain treshold? Like for example, in my current setup
(client side), everything with a spam score of .99 gets moved to /dev/null
by procmail. On the postfix server level, I would like to fail emails with
a spam score of .99 with an SMTP error, like for example error 554:
Transaction failed.
Is that possible, and if yes, how can I do that?

I might have some other questions as I go along, but let's start with these.

-- 
Amedee Van Gasse
amedee at amedee.be



More information about the SpamBayes mailing list