RE: getting spamassassin learn to separate spam from ham, postfix and exchange

[Robert Bartlett <ro...@digitalphx.com>]

Perhaps you can use fetchmail on the linux box? With some help I was able to
create a fetchmail script that did the following:

1. Look for account info for fetchmail via a mysql database
2. Grab the email from that account using fetchmail
3. Feed it to SA on the linux box and filter it like a normal email to the
linux mailbox

That way you can use your normal filters, such as maildrop, to determine if
it is spam or ham and move them to the appropriate location based on the
scoring. Perhaps you could create a single mailbox where the email gets
delivered to from the fetchmail script and have a ham and spam folder. Then
if it is spam it goes into the spam folder and visa versa for ham. That way
the majority stuff is done automatically. I just setup the fetchmail script
as a cron that checks every 15 minutes.

Robert

-----Original Message-----
From: Mike Sassaman [mailto:msassaman@strategictele.com] 
Sent: Thursday, January 12, 2006 9:09 AM
To: users@spamassassin.apache.org
Subject: RE: getting spamassassin learn to separate spam from ham, postfix
and exchange

I would also be interested in the answer to this.  Although I am using
Sendmail + smtpvilter + SA (with spamd) as a relay, with an Exchange server
behind it.  

Basic questions is the same: How do I get spam off the exchange server en
masse to train bayes?

-----Original Message-----
From: Karsten Krämer [mailto:kk@california3000.de]
Sent: Thursday, January 12, 2006 4:32 AM
To: users@spamassassin.apache.org
Subject: getting spamassassin learn to separate spam from ham, postfix and
exchange


Hi,

we use postfix, amavisd-new, spamassassin on a linux box as relay, exchange
on server 2003 as final destination.

Which way is the best to get (ham and/or spam) mail out of user mailboxes
from exchange to linux for further "sa-learning".
Maybe there exist some Howto - we didn't manage to find them yet.

At the moment we collect the body texts in two big files and manually learn
spamassassin to separate the "good from the ugly", but in this way all
headers are lost.

Any more advanced ideas? Any help appreciated.
Thanks

Karsten