On 3/21/06, Gary W. Smith <[EMAIL PROTECTED]> wrote: > It appears that you are user per user bayes. If you have a large number > of users and performance is becoming an issue you might want to change > over to site wide bayes and disable per user.
How are you training Bayes? Do you allow the users to submit emails to train the system? Any chance of rogue users screwing things up that way? > As for user purging, no, there is no user detection to process users > that are no longer in the system. Ok, so if I understand correctly, users that no longer exist are still in the database taking up space? Time for some code foo... -- Jason 'XenoPhage' Frisvold [EMAIL PROTECTED]