Martin, > > I just let it start from scratch, depending entirely on autolearning > > from SpamAssassin. This only works well if your SpamAssassin is producing > > quality results (uses good network tests, updated rules, > > sought.rules.yerp.org > > One advice: start with a very low score, so the classifications will be > visible in the headers but cannot cause false positives.
Definitely! > Do you have the impression that dynamic scores give better results than > static scores? > > Since I did not see an advantage on my setup and there is no documented > SpamAssassin API to set them, I even thought about removing the dynscore > option. I don't have a strong proof. I started with static rules, but then it seemed to me the dynamic score with its continuous (non-discrete) score might be useful and I switched to dynamic score, which serves me well, so I never looked back. A high score from crm114 when it is very much sure about its spam verdict can be useful to bump the overall score above $sa_dsn_cutoff_level, but then again, I had an instance of two when its high score was wrong. Seems both approaches are valid. What counts most is to keep an eye on results, and intervene when necessary (at the SA or at the crm side). Mark ------------------------------------------------------------------------------ Crystal Reports - New Free Runtime and 30 Day Trial Check out the new simplified licensign option that enables unlimited royalty-free distribution of the report engine for externally facing server and web deployment. http://p.sf.net/sfu/businessobjects _______________________________________________ AMaViS-user mailing list AMaViS-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/amavis-user AMaViS-FAQ:http://www.amavis.org/amavis-faq.php3 AMaViS-HowTos:http://www.amavis.org/howto/
