Have now installed the non-forwarding dns server unbound and am no longer black listed, yay! Given that spamassassin and amavisd both use the same underling Perl modules, I still don't fully understand why they are giving such wildly different results. Still, it seems that things have improved so your advice & help are much appreciated.
Will On 31/08/17 20:56, Benny Pedersen wrote: > Will Hall skrev den 2017-08-31 21:41: >> Many thanks for the reply. > > +1 > >> Not sure I completely follow when you say "spammers have a free ride on >> your limited data'. > > its possible not just uribl, but uribl tells you need to solve dns > >> Yes, I am blocked by uribl, but resolving this is not trivial (am >> looking into it). However, I don't think this explains why the bayes >> scores are so different - are the two issues connected? > > good, bayes track changes, and raw amavisd and spamassassin can get > diffrent results, its not a bug anywhere, if you get dns problem solved > bayes will learn more on bad or good things, so you dont need to train > bayes so much anylonger, all in all solve dns and report problem after > you have this working
