http://bugzilla.spamassassin.org/show_bug.cgi?id=2094
------- Additional Comments From [EMAIL PROTECTED] 2004-12-02 05:17 ------- Subject: Re: Proposal: New autolearn variant > IMHO using the bayes score to determine autolearning isn't a good > idea at all. It'll only help to push a partially 'poisoned' bayes > database further into worthlessness. This is only true if spamassassin is not supervised at all. If you check filtered mails for false pos/neg the effect will be reversed: the bayes db will use it's own data to improve itself. I described this in the original proposal. --Nikolaus ------- You are receiving this mail because: ------- You are the assignee for the bug, or are watching the assignee.
