On Wed, 2004-03-24 at 15:49, Stephen Smoogen wrote: > Ok here is what I think happened for me at least.. I didnt have enough > SPAM in my bayesian filter as the sa-learn --dump magic only showed 155 > spam messages. I thought I had feed it a lot more but I think that it > had 'forgotten' some. I noticed that a lot of tests have a weight of 0 > when Bayes is turned on so I think spamassassin was not giving it the > full weight it should. >
Yes, just learn more spam. Actually, I have difficulty understanding why I would want to use bayes at the user level because most folks either do not have the knowledge or do not have the time to gain the knowledge that is needed to handle their own bayes database. There are probably some sophisticated tools available commercially, but doesn't spam cost enough already! My site-wide bayes has over 4000 ham and over 4000 spam and it is extremely rare for a spam message to get through undetected. Granted my volumes are low, only 170 or so spam messages a day and I may see 1 or 2 spam messages a month in my inbox. I think you mentioned Can-it, I have not seen that one but I do know that SA rocks :) Alex _______________________________________________ Visit http://www.mimedefang.org and http://www.canit.ca MIMEDefang mailing list [EMAIL PROTECTED] http://lists.roaringpenguin.com/mailman/listinfo/mimedefang
