>I shut off "bayes_auto_expiry" and then do the forced expire at night. I don't control this from MS. You may want to up the expiry threshold to 500.000 or so (I think the default is 100.000).
So, did you set "Rebuild Bayes Every = 0" in your MailScanner.conf? Is it better if both this setting and the "bayes_auto_expire" in spam.assassin.prefs.conf are set to 0? > It doesn't matter if you learn before or after expire. (well, it effects > the number of removed tokens a bit, of course, but new tokens won't be removed, anyway). If you learn to the journal you also need to --sync or the just learned stuff won't be of use for a while. Since --force-expire should also force a sync with the journal, I was wondering if it would be better to run the force-expire after having learned from the daily spam folder... > These are the remnants from your earlier timeouts. If you ever encounter these it means that the expiry doesn't finish. Thanks for the tip, I've deleted them... -- View this message in context: http://www.nabble.com/Some-questions-on-sa-learn-tp21541965p21546134.html Sent from the SpamAssassin - Users mailing list archive at Nabble.com.