----- Original Message ----- From: "Kris Deugau" <[EMAIL PROTECTED]>
> > That would preserve all the headers but I didn't > > know if sa-learn would know to just look in the attachment or not? > > Not automagically. You'd have to build a script to pull out the > attached message and sa-learn *that*. I assume that's been done before? Are there any scripts floating around to do that? > I would recommend against blindly letting users feed Bayes unless > they're just feeding their own individual Bayes databases- *especially* > in an ISP environment. This is just a fairly large company (not an ISP). Right now it's just a common Bayes for everyone. How would I setup qmail+qmail-scanner to work on an individual bases? Also how would you deal with setting up addresses that each user can forward their spam or ham to? Something like .qmail-spam-default and then users send spam to [EMAIL PROTECTED] and then a script in .qmail-spam-default parses out theirusername and updates the correct bayes? I guess? Of course what stops users from sending spam to [EMAIL PROTECTED] and messing with each other? Thanks Malcolm
