I would like to teach my spamassassin instance using sa-learn, however I'm not sure how to feed spam to it.
With sa-learn you can feed it an mbox or mbx in which you've put spam messages, however, since using vpopmail, I don't use the native mail folders and since all mail is popped, there's none left of the server to learn from anyhow!
Has anyone else tried to use sa-learn for bayesian learning?
Is it possible that when spamassassin detects a spam message, aside from modifying the subject header, it also emails a copy to the local machines mail account (one of the system accounts) and run sa-learn against it?
Cheers,
