Hello all,
I started to get increasingly annoyed by certain spam mails getting
through my previous 2.63 version of Spamassassin, mainly because the
Bayes filter didn't bite any more, so I had this weird idea about
upgrading my two-year-old thing with a brand new one. So I did. After
some slight annoyances of downloading a few modules from CPAN and
installing them (no problems, just some work) I got the new thing up and
running smoothly.
The only problems is that I see no improvement. In particular, the Bayes
filtering remains as misled as it was previously, possibly even worse.
So I decided to remove the bayes_* files from the respective directory
to start it all over again. I fed sa-learn utility with 1000 mails of
ham and spam, and hoped for good. And no, I didn't get the flags wrong,
neither did I mix the messages. Honestly.
But still I have the Bayes filter declaring very typical spam messages
(gambling words, drugs etc) as BAYES_00, which means "surely ham". Not
to mention those mails in which the name of the drugs are in capitalized
letters with spaces between them, and then a paragraph with "normal text".
I'm pretty close to giving up the Bayes filter, and let Razor2 do the
job instead. Before I do that, does anyone here have a comment to make
on this? Are spam messages becoming smarter with Bayes filters, or...?
Thanks,
Eli
--
Web: http://www.billauer.co.il
--------------------------------------------------------------------------
Haifa Linux Club Mailing List (http://www.haifux.org)
To unsub send an empty message to [EMAIL PROTECTED]
