Benny Pedersen wrote: > On Sun, September 24, 2006 04:50, Matt Kettler wrote: > >>> change score on whitelist_from solves most here >>> >> Eh? How does changing the score solve anything? >> > > we talked about usefullness on whitelist_from ? > > if changing score can make it better since its the way spamassassin works on > > no one wants to have whitelist_from score -100 > I know how it changes things. I asked how it SOLVED anything.
You can easily change the score for USER_IN_WHITELIST, but I don't see the point. What other score would you use? -20? -10? -5? -1? 0? +1? What actual real advantage is there to this score compared to the default? > >>> else ham is not learned from local users >>> >> You do know that the whitelist score doesn't affect the autolearner, right? >> > > awl do use the scores from whitelist_from ?, i just make sure its awl > autolearn not bayes > > bayes does not > First: Neither the AWL nor bayes autolearning should be affected by the scores from whitelist_from. The USER_IN_WHITELIST rule is declared: with these tflags: userconf nice noautolearn. Both the AWL and bayes autolearner will ignore any rule with the "noautolearn" tflag. Also, beware it's a slight misconception to call it "AWL autolearn". The AWL learns EVERY message. There are no autolearn thresholds involved, like there are for bayes.
