Bayes not working

[Robert Leonard <RL...@Opto22.com>]

 I've deleted all my bayes databases and am trying to start the database
from scratch.. I auto-learn..

For some reason SA sees 8 spams in the database but won't learn any more...

What files does bayes need?  I have the bayes_toks and bayes_seen files, but
nothing more..

Thanks all!

Below is the part of the lint that seems to imply it is attaching...  But I
have received numerous spams scoring above my 20 point threshold since
starting up SA this morning..  

------------------------------------------------------------------------

debug: bayes: 29956 tie-ing to DB file R/O /bayes/bayes_toks
debug: bayes: 29956 tie-ing to DB file R/O /bayes/bayes_seen
debug: bayes: found bayes db version 3
debug: bayes: Not available for scanning, only 8 spam(s) in Bayes DB < 200
debug: bayes: 29956 untie-ing
debug: bayes: 29956 untie-ing db_toks
debug: bayes: 29956 untie-ing db_seen
debug: Score set 1 chosen.
debug: ---- MIME PARSER START ----
debug: main message type: text/plain
debug: parsing normal part
debug: added part, type: text/plain
debug: ---- MIME PARSER END ----
debug: bayes: 29956 tie-ing to DB file R/O /bayes/bayes_toks
debug: bayes: 29956 tie-ing to DB file R/O /bayes/bayes_seen
debug: bayes: found bayes db version 3
debug: bayes: Not available for scanning, only 8 spam(s) in Bayes DB < 200
debug: bayes: 29956 untie-ing
debug: bayes: 29956 untie-ing db_toks
debug: bayes: 29956 untie-ing db_seen

Re: Bayes not working

[Matt Kettler <mk...@evi-inc.com>]

At 01:49 PM 9/29/2004, Robert Leonard wrote:
>  I've deleted all my bayes databases and am trying to start the database 
> from scratch.. I auto-learn..
>
>For some reason SA sees 8 spams in the database but won't learn any more...
>
>What files does bayes need?  I have the bayes_toks and bayes_seen files, 
>but nothing more..

That's all it needs. Sometimes it makes a journal file, but that eventually 
gets re-synched into the toks file and deleted.


>Below is the part of the lint that seems to imply it is attaching...  But 
>I have received numerous spams scoring above my 20 point threshold since 
>starting up SA this morning..

Just because the email scored 20 points doesn't mean it's above a learning 
threshold of 20 points.

Bear in mind that autolearning isn't based on the final email score, it's 
based on a recomputed score that's calculated with bayes disabled, 
including scoreset change, and with any userconf rules disabled. It also 
won't autolearn anything as spam unless it has 3.0 worth of header AND 3.0 
of body, regardless of total score. It also tries not to learn anything 
that strongly contradicts existing learning.

Really, I'd shy away from auto-learn only if possible. Even if you only 
manually train a small amount of mail at the start and let it autolearn 
from then forward you're likely to get a much better bayes database.