On Sat, 26 May 2012 22:44:54 +0200 Wolfgang Zeikat wrote: > In an older episode, on 2012-05-26 22:38, Wolfgang Zeikat wrote: > > > We had so many false positives > > Oops, I used your term "false positives" by accident. I and many > others tend no call false Ham classifications > false negatives > (negative scores change the classification towards ham)
It depends on context. He was originally wrong because he wrote that he was getting false positives in his inbox, which implies a false positive in the overall spamassassin result. OTOH RCVD_IN_DNSWL_MED hitting spam is a false positive in the individual rule which is a test for ham. > So: > We had so many false negatives > > with that rule, that I - as others who > > replied to your post already (see below) - have come to the > > conclusion that www.dnswl.org is not a reliable source of trust for > > us and disabled the rule by configuring > > > > score RCVD_IN_DNSWL_MED RBL 0 The OP should probably update his rules first since the rule currently scores -2.3 rather than -4, and rules haven't been updated since February. I don't think setting it to zero is a good idea, it wont turn-off the lookup so you might just as well set it to -0.001 and monitor the rule's performance.
