I guess you would do checks on the negative weights first and then the positive and at any point a test goes above the threshold you would stop. Unless by adding all the positive tests together it would still be below threshold whereas you wouldn't need to do any positive tests (unlikely though).
This would follow from your suggestion that there are always more positive than negative tests, so do the few before the most. Also I'd do the positive DNS tests last as they are the ones that take the most time, so they may not even be needed. If you are multithreading you could fire off all the DNS tests and then do the other tests while you are waiting for responses? Jools On Sun, 4 May 2003 09:09:45 -0400, you wrote: >Interesting point, I hadn't considered tests that add a negative weight. >(Although the default config only comes with two negatives as far as I know >and I haven't had need to add others). However I think it would be safe to >make some assumptions in the test process that could cut CPU time and still >yield the type of accuracy we're used to. For instance, its very likely >you'll have more tests that add to the weight than those which detract. So >you could skip the detracting tests on those messages which haven't gathered >enough weight from the adding tests to fail your threshold. Why bother if >they're going to pass anyway right? Along the same line of thought, you >could just stop processing additive tests after your weight threshold has >been met and simply configure the detracting tests weight values according >to your threshold. > >Thoughts? > >Rob Salmond >Ontario Die Company >(519)-576-8950 ext. 132 > > >--- >[This E-mail was scanned for viruses by Declude Virus (http://www.declude.com)] > >--- >This E-mail came from the Declude.JunkMail mailing list. To >unsubscribe, just send an E-mail to [EMAIL PROTECTED], and >type "unsubscribe Declude.JunkMail". The archives can be found >at http://www.mail-archive.com. --- [This E-mail was scanned for viruses by Declude Virus (http://www.declude.com)] --- This E-mail came from the Declude.JunkMail mailing list. To unsubscribe, just send an E-mail to [EMAIL PROTECTED], and type "unsubscribe Declude.JunkMail". The archives can be found at http://www.mail-archive.com.
