Volker Kuhlmann wrote:
The newer mozilla builds have bayesian spam filtering: You teach it what
spam looks like and it filters automatically.
Reading the instructions for the spamassassin-bultin bayesian
filtering, it seems it's not so straightforward. As usual, the
sure-fire spam killer doesn't exist. How involved is "you teach it"?
For junk mail not identified as such, you press the "junk" button.
For real mail identified as junk, you press the "not junk" button.
(Rhetorical question, I'm not prepared to depend on some huge bloatware
for mail filtering so won't use it anyway.)
Doesn't worry me, I've got plenty of RAM. Mozilla is only taking up
62MB of it :-)
I got >120 spams over the weekend. It caught every one. No false +ves
and no false -ves.
Hm. Despite what people rave about, even after tweaking (increase html
scores, add those patterns, rbl lookups, etc) I only reach a 95-98% hit
rate with SA. No false +, but spam specifically crafted to slip through
SA only scores about 2.5 (I've seen several of them now).
Just after sending my previous email I *did* get a spam which was not
identified as spam - in some foreign language :-( This only happens
maybe 1-3 times per week. Simply clicking on the "Junk" button means
I'll never get that spam again.
For ease of use, it beats explicit rule-based systems (i.e.
spamassassin) hands down IMHO.
Cheers,
Carl.
