On 3/11/02 9:29 PM, "Forrest Cahoon" <[EMAIL PROTECTED]> wrote:
> On 03/11/2002 20:30:17, Michael Moncur wrote: > > (Re: "spam" rules with negative weights) > >> Apparently these rules are either no longer significant spam indicators or, >> in extreme cases, have become non-spam indicators. > > The mind boggles. How could the presence of "$$$" carry any weight > _against_ an e-mail being spam? I understand the algorithm ran > against a large statistical sample, but perhaps some rules should not > be allowed to go negative (or positive, in the case of obvious > non-spam rules), leading to an adjustment of the other numbers. Sometimes maybe people use lines of $'s as separators, I don't know. It's also possible that in instances where the mail is spam containing $'s that there are lots of other rules being triggered, and those are enough to push the message over the threshold, so the $ rule is scored lower. > Where can I find out more about the weighting algorithm? I haven't > come up with anything poking around spamassassin.org (but I have been > known to miss the obvious before). In the masses directory of the distribution. The craig-evolve.cxx is the one currently being used. The README file has some info, but the code is easy to read. > (Re: rule for meta tag indicating MS FrontPage) > >> Didn't someone already add a rule for this? > > It's not in the version 2.11 production release, or at least I did a > case-insensitive grep of the rules directory for "meta" and > "frontpage", and both searches turned up empty. Annotations for rules/20_body_tests.cf *************** 1.54 (hughescr 05-Mar-02): rawbody FRONTPAGE /FrontPage.Editor/ 1.54 (hughescr 05-Mar-02): describe FRONTPAGE Frontpage used to create the message Annotations for rules/50_scores.cf *************** 1.44 (hughescr 05-Mar-02): score FRONTPAGE 2.000 _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
