Mike Jackson wrote:
>> TOP HAM RULES FIRED
>> ------------------------------------------------------------
>> RANK    RULE NAME                       COUNT %OFRULES %OFMAIL %OFSPAM
>> %OFHAM
>> ------------------------------------------------------------
>>   1    DNS_FROM_RFC_ABUSE              88943    13.50   15.85   12.68
>> 25.27
> 
> That worries me. Granted, at most that will add 0.479 to a message (when
> used with the default 3.1.1 scores), but to have your #1 *ham* rule be
> one that's supposed to identify *spam* doesn't speak well for the rule.
> I like RFCI; I feed it bogusmx or DSN-violating mail whenever I can.
> But, the abuse and postmaster lists contain far too many *major* ISPs
> for them to be reliable indicators of spam.

SpamAssassin's perceptron appears to agree with you, it has never in history
given this rule as much as 0.5 points:


Mail-SpamAssassin-3.0.0/rules/50_scores.cf:score DNS_FROM_RFC_ABUSE 0 0.374 0 0
Mail-SpamAssassin-3.0.1/rules/50_scores.cf:score DNS_FROM_RFC_ABUSE 0 0.374 0 0
Mail-SpamAssassin-3.0.2/rules/50_scores.cf:score DNS_FROM_RFC_ABUSE 0 0.374 0 0
Mail-SpamAssassin-3.0.3/rules/50_scores.cf:score DNS_FROM_RFC_ABUSE 0 0.374 0 0
Mail-SpamAssassin-3.0.4/rules/50_scores.cf:score DNS_FROM_RFC_ABUSE 0 0.374 0 0
Mail-SpamAssassin-3.1.0/rules/50_scores.cf:score DNS_FROM_RFC_ABUSE 0 0.479 0 
0.200
Mail-SpamAssassin-3.1.1/rules/50_scores.cf:score DNS_FROM_RFC_ABUSE 0 0.479 0 
0.200

Reply via email to