>However a way around this has just occured to me: > >1. Run all negatively scoring tests. > >2. Run positively scoring tests in highest-score first order. > >3. Stop when we hit the threshold.
How about: 1) run the positive tests until it hits the threshold 2) run all the negative tests 3) run the remaining positive test until it hist the threshold again And I think the test positive tests should be run in order of relevance rather than in order of costs maybe. That is the tests that trigger the more often are run first (a strong indicator of spam, that would cost +10 is not usefull to run if it ever catched a single mail from the corpus). Olivier _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
