On Tue, 17 Mar 2009 23:41:23 +0100 fl...@pbartels.info wrote:
> instead of disabling a lot possibly set message headers using > "bayes_ignore_header" and ending up in strange configs like: > > <snip header list> > > (found on the net) Just because you found them on the net doesn't mean it's a good idea. Normally you avoid learning upstream antispam headers because you want the BAYES result to be independent, i.e. you don't want to train it to fail in line with the non-bayes rules. > shouldn't SpamAssassins bayes mechanism just ignore the complete > message header and just look at the body? Headers can be a very useful source of information. In particular they allow BAYES to learn spam-free deliver paths, enabling ham to hit BAYES_00 even on very short messages, thereby helping to avoid FPs.
