Stops after RCVD_IN_BL_SPAMCOP_NET

Kai Schaetzl maillists at conactive.com
Fri Jan 9 22:31:19 GMT 2009


Joe Garvey wrote on Fri, 9 Jan 2009 11:32:49 -0800:

> My /etc/mail/spamassassin/spam.assassin.prefs.conf is a link to /etc/MailScanner/spam.assassin.prefs.conf

right, yes, that's fine then. I sometimes think we are two years ago ;-)


> I find it very confusing and lacking confidence in the system when
> the system provides a score for bl.spamcop.net and don't see any other
> results from any other rules. 

Use Mailwatch to checks the Rule Hits. Go to Reports/Spamassassin Rule Hits.
I can't see a reason that the spamcop RBL rule stop all processing. Unless you use
short-circuiting and use this rule as short-circuit rule.

> I also converted my bayes database to MySQL. After reviewing the conversion
> I noticed that I have no ham messages in the database. I am loading
> up some from various users to see if this will also make a difference
> as I find the Bayesian score usually shows a negative even for the
> most obvious spam.

Well, I suppose you didn't do "sa-learn --dump magic" before this. It would have
shown you that you have no ham. That symptom would be normal for a freshly started
Bayes DB that gets trained only with autolearning, but you seem to have it running
much longer? This would then indicate that your autolearning for ham is non-existent
because your ham isn't scoring low enough - which should not happen.
What's the dump magic output now?

Kai

-- 
Kai Schätzl, Berlin, Germany
Get your web at Conactive Internet Services: http://www.conactive.com





More information about the MailScanner mailing list