SA and Bayes

[Julian Field]

At 14:03 06/03/2003, you wrote:
>What happens if you don't specify anything for bayes and go with the
>defaults?  I think my setup is using bayes because the .spamassassin dir
>under root has bayes files that are updating.  The only problem that
>I've been seeing with bayes is that it's not very good at cleanup.  It
>leaves lock files there but I've been noticing more sa temp files being
>left behind in /tmp as well.

But have you actually seen "BAYES" in any of the spam reports? Switch "Log
Spam = yes" and sit back and wait. I've yet to see this being used in
anger, although the db files are indeed being updated.

I'm currently trying to recreate a faulty system so I can test this. Just
got to copy my 60,000 message test set onto the poor thing :-)


>-----Original Message-----
>From: Jan-Peter Koopmann [mailto:Jan-Peter.Koopmann at SECEIDOS.DE]
>Sent: Thursday, March 06, 2003 3:36 AM
>To: MAILSCANNER at JISCMAIL.AC.UK
>Subject: SA and Bayes
>
>
>Hi,
>
>somehow I get the impression that my SA/MS setup does not use bayes.
>These are the relevant entries of my spam.assassin.prefs.conf file:
>
>auto_whitelist_path        /var/spool/spamassassin/auto-whitelist
>auto_whitelist_file_mode   0666
>
>bayes_path                 /var/spool/spamassassin/bayes
>bayes_file_mode            0666
>
>auto_learn 1
>
>use_bayes 1
>
>bayes_ignore_header X-MailScanner
>bayes_ignore_header X-MailScanner-SpamCheck
>
>
>This is what a check_bayes_db tells me:
>
>0.000        0        0        0  non-token data: db format = on-the-fly
>probs, expiry, scan-counting
>0.000        0      209        0  non-token data: nspam
>0.000        0     2320        0  non-token data: nham
>0.000        0        0        0  non-token data: ntokens
>0.000        0        0        0  non-token data: oldest age
>0.000        0     4505        0  non-token data: current scan-count
>0.000        0        0        0  non-token data: last expiry scan-count
>
>--- snipp ---
>
>If I interpret this correctly I have 209 spam and 2320 nonspam messages
>learned successfully. And as far as I can see both sa-learn and
>auto_learn seem to work. BUT: I never saw a single mail (spam and
>nospam) with a BAYES_ tag in the SpamAssassin score.
>
>Any ideas?
>
>Regards,
>    JP

--
Julian Field
www.MailScanner.info
MailScanner thanks transtec Computers for their support