It is a bit difficult now to continue on a thread that doesn't exist anymore, but i'd like to recreate this thread for people to share best practices on controlling SPAM and setting up the Scores properly in SPAMASSASSIN.
In the old thread, I had a user by the name of richard (he still has to join the new forums again though) who discussed some methods in using the "sa-learn" command in SSH to teach the bayesian filters about uncaught spam.
If i remember correctly, the method he described was to create a folder under your mailbox called "uncaught" or something similar, and moving all SPAM messages that slipped through spamassassin in this folder.
The, to make SA aware of this, you run the command:
sa-learn -D --mbox --spam ~/mail/yourdomain.com/yourmailbox/uncaught
After running this command from SSH, SA will know about these messages that it let through, and eventually learn to mark these with a higher spam score.
richard can run this command on his bluehost account with no problems, but unfortunately, when I run this command on box77, I get some errors:
Can anyone else confirm if this is working on their accounts, and how to enable the particular DB_File module?
 dbg: metadata: X-Spam-Relays-Trusted:
 dbg: metadata: X-Spam-Relays-Untrusted:
 dbg: message: ---- MIME PARSER START ----
 dbg: message: main message type: text/plain
 dbg: message: parsing normal part
 dbg: message: added part, type: text/plain
 dbg: message: ---- MIME PARSER END ----
 dbg: message: no encoding detected
 dbg: bayes: DB_File module not installed, cannot use bayes
Learned tokens from 0 message(s) (1 message(s) examined)
ERROR: the Bayes learn function returned an error, please re-run with -D for more information
And continue sharing new ideas please