> Simon Byrnand <[EMAIL PROTECTED]> wrote: >>I wonder if anyone has any stats or anecdotal evidence on the relative >>effectiveness of Razor2, DCC, and Pyzor ? > > I don't have any stats, just anecdotal evidence. I've found Pyzor to be > effective, sometimes more so than Razor, sometimes less. Certainly when > spam appears in both databases, it's almost guaranteed to be caught.
Well I tried turning on all three yesterday, and collected stats on spam that score between 5 and 15, and so far I've got: RAZOR2 - 383 DCC - 225 PYZOR - 121 Out of a total of 992 messages. (Remember this excludes message that scored 15 or above) > I haven't used DCC, mainly because its purpose is different. It > deliberately includes legitimate newsletters because it doesn't measure > how > spam-like a message is, but how many people have received it. Since I > manage a sitewide setup, it just seems like it would be too much effort to > go through and whitelist every newsletter received by every user. Hmm, you've got me a little bit worried there, but where do you get the idea that DCC "deliberately includes legitimate newsletters" ? I've seen newsletters listed in RAZOR2 quite often, although they sometimes later get revoked. And when you say DCC doesn't measure how "spam-like" a message is, I wasn't aware that Razor2 or Pyzor did either ? All three rely on submissions from people to add to the database, I don't see how the DCC/RAZOR2/PYZOR databases can (or should) try to analyze the message content to see if its spam or not, which would be duplicating the efforts of programs like SpamAssassin for one... Despite the fact that DCC seems to advocate the use of automated spamtrap submissions while RAZOR2 discourages it, one very telling factor is that the GA has scored DCC quite a bit higher than either RAZOR2 or PYZOR, which suggests the GA has detected less false positives than the others, at least in the corpus(es) available for the GA to run on.... Certainly my initial impression after 24 hours is that DCC is the most effective, followed by RAZOR2, followed by PYZOR. All together is probably most effective of course, so long as the false positives of all three dont accumulate enough to register as spam. > One thing I've found helpful with both Razor2 and Pyzor is to keep the > timeouts short in local.cf. That way if some error comes up, it doesn't > delay the mail too much. Yep, I've got all three, and the RBL checks set to timeout after 10 seconds... ------------------------------------------------------- This SF.Net email sponsored by: Parasoft Error proof Web apps, automate testing & more. Download & eval WebKing and get a free book. www.parasoft.com/bulletproofapps _______________________________________________ Spamassassin-talk mailing list [EMAIL PROTECTED] https://lists.sourceforge.net/lists/listinfo/spamassassin-talk
