Loren, Andreas, thanks for such a fast reply. I always urge people to give complete information when asking or reporting bugs but don't do it myself :( I am sorry!
I use the copfilter (http://www.copfilter.org/) plugin 0.83 beta3a to ipcop firewall which contains SpamAssassin 3.1.4. Loren seems right - the spamd log file Andreas reminded me of says bayes=0.494975644302803: Oct 8 09:05:08 SpAs spamd[10825]: spamd: connection from localhost [127.0.0.1] at port 38360 Oct 8 09:05:09 SpAs spamd[10825]: spamd: processing message <[EMAIL PROTECTED]> for filter:702 Oct 8 09:05:26 SpAs spamd[10825]: spamd: identified spam (9.2/5.0) for filter:702 in 17.5 seconds, 10639 bytes. Oct 8 09:05:26 SpAs spamd[10825]: spamd: result: Y 9 - DATE_IN_FUTURE_03_06,EXTRA_MPART_TYPE,HTML_40_50,HTML_IMAGE_ONLY_12,HTML_MESSAGE,SPF_HELO_SOFTFAIL,SPF_SOFTFAIL scantime=17.5,size=10639,user=filter,uid=702,required_score=5.0,rhost=localhost,raddr=127.0.0.1,rport=38360,mid=<[EMAIL PROTECTED]>,bayes=0.494975644302803,autolearn=no Oct 8 09:05:26 SpAs spamd[10556]: prefork: child states: II Training the mails does not help too much: it is the type with random text and an embedded picture. I have no idea how to fight them except lowering the SPAM score. Thanks again both of you, Paul -- View this message in context: http://www.nabble.com/sometimes-no-bayesian-filtering--tf2404032.html#a6702261 Sent from the SpamAssassin - Users mailing list archive at Nabble.com.
