-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1
Eugene Morozov writes: > > We'd like to add support for it. Anyone fancy coding it up? > > If you read the comments at slashdot, there's a link to > SpamAssassin bugzilla where you can find patch for > integrating some old version of SA (2.60 or older) with > crm114. But it seems from the discussion of the bug > that SA developers doesn't want to integrate crm114 or > something like that with SA because they have enough > headache with built-in bayes. They recommend to use a > new "plugin" feature to implement such things. I'll > look into this and will try to integrate SA with crm114 > (which I'm evaluating now) or/and DSPAM (which I'm also > going to install and test). BTW, what I was talking about was the Dobly noise reduction technique, not DSPAM itself. > As to anti-SA comments, he's partially right. Rules > only make sense when someone constantly updates them and > carefully choose scores. So, there must always be > someone who creates new rules, collects rules from the > web (maybe using such tools as Rules Du Jour), install > and test them. This is rather cumbersome and > ineffective. there's no doubt that a manually-built rule-based system requires someone to collect and write the rules. The anti-SpamAssassin stuff I didn't like was more regarding how crappy perl is, how SpamAssassin requires more training and tweaking than a pure-bayes filter does because it's so inaccurate, some FUD about how DSPAM infers the same stuff as SpamAssassin's rules do by picking an extremely simple example which isn't in the distro anymore, etc. etc. etc. - --j. -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.2.4 (GNU/Linux) Comment: Exmh CVS iD8DBQFAO8mRQTcbUG5Y7woRAvA6AKDosLDA/on1qDDMxuClwXjm8sT5HwCeNdoq 7kFhzqdE7xVCgKHd4evx8WY= =J+X+ -----END PGP SIGNATURE-----
