Matus UHLAR - fantomas wrote: >> On Thursday 11 October 2007, Mark wrote: >> >>> I'm new to the list, so I hope this is the right place. >>> >>> I am running my mail through procmail and separating my spamassassin >>> into 3 groups depending on score: >>> >>> X-Spam-Status: Yes, score=[2-9][0-9] >>> X-Spam-Status: Yes, score=[0-9][0-9] >>> X-Spam-Status: Yes >>> >>> I reason that this must be in order of confidence. I have found that the >>> first group (score>=20) contains more than half of my spam. The last >>> group (score<10) contains about 1/10 of the spam. >>> >>> Would others agree that I can safely get procmail to trash the scores >>> higher than 10? >>> > > On 11.10.07 05:23, Gene Heskett wrote: > >> I've been sending that stuff to /dev/null at >7, no false hits so far >> according to the procmail.log. My bayes is fairly well trained. >> > > We (ISP I work for) had complaints when the messages scored over 7 were > refused. I have even seen some messages that could be taken as false > positives scoring a 11. > > However I take refusal with score over 10 as safe. Some people just have to > fix their submission agents (usually PHP scripts who don't care about proper > encoding and splitting lines) > >
refusal (at smtp time) is ok. but OP is about discarding. This is generally less acceptable. If it's one's mail, then one is free to do whatever he wants. but if it's other people's mail, he should ask them (after explaining the risks).
