I think one thing we're learning with each x.y0 release of spamassassin is that rule scores need to be tweaked after the GA runs, and that within a week or so after x.y0 we need to release x.y1, which fixes almost all scoring issues. I agree that 4.1 is probably a little high for that rule; probably what I should do is sign up a honeypot-like address to a bunch of legitimate mailing lists, and then use that as part of the nonspam corpus, after manually removing any spam sent to the lists. That would probably yield great improvements in the scoring accuracies for those types of patterns.
C Klaus Heinz wrote: KH> Hi, KH> KH> I've been using SA for about 2 months now and have been running with the KH> default threshold of 5 hits. KH> KH> With the new version 2.20 I got a false positive with a newsletter KH> I receive. KH> KH> X-Spam-Status: Yes, hits=6.8 required=5.0 tests=EXCUSE_3, KH> HTTP_WITH_EMAIL_IN_URL version=2.20 KH> X-Spam-Report: 6.8 hits, 5 required; KH> * 2.7 -- BODY: Claims you can be removed from the list KH> * 4.1 -- URI: 'remove' URL contains an email address KH> KH> The point is not, how to avoid SA marking this specific newsletter KH> as spam, I can do that, but the score of 4.1 for HTTP_WITH_EMAIL_IN_URL KH> seems a bit high. KH> SA 2.11 scored this mail with 1.0 and 1.8 (and also matched PENIS_ENLARGE2 :-). KH> KH> Or is a threshold of 5 too low ? What do other people use ? KH> KH> Can I report false positives as I do with unrecognized spam to KH> [EMAIL PROTECTED]? KH> KH> ciao KH> Klaus KH> KH> KH> _______________________________________________ KH> Spamassassin-talk mailing list KH> [EMAIL PROTECTED] KH> https://lists.sourceforge.net/lists/listinfo/spamassassin-talk KH> KH> KH> _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
