On Tue, 2 Sep 2003, Ron Gilbert wrote: > I guess my complaint is that I would have rather it started working and > been somewhat useless (like POPFile did), or provide much better > feedback on it's status.
SpamAssassin's ethos has always been that a few false negatives are better than even one false positive. Always err on the side of not tagging, as it were. Hence, don't believe in Bayes until it has been well-trained. > From what I can tell, it really isn't *200* ham/spams, it's 200 > different enough ham/spams. "Different enough" just means that they have different message-id headers, at least in 2.5x. I haven't been watching whether this has been revised for 2.60. Are you actually feeding it spam and ham with "sa-learn", or are you just relying on auto-learning? ------------------------------------------------------- This sf.net email is sponsored by:ThinkGeek Welcome to geek heaven. http://thinkgeek.com/sf _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
