More spam after spamassain upgrade

John Rudd jrudd at UCSC.EDU
Sat Jul 26 00:27:42 IST 2003

On Thursday, Jul 24, 2003, at 16:16 US/Pacific, Matt Kettler wrote:
> It's also pretty well known that the ruleset in the 2.5x family is
> considerably weaker than the later part of the 2.4x family. But the
> ruleset
> weaknesses are largely made up for by the power of bayes, and are
> mostly a
> result of the developers focusing on getting bayes added instead of new
> rules while writing 2.50. If you're going to use 2.5x, train your
> bayes DB,
> it helps a lot.

I wonder when this is going to be resolved.  They haven't released 2.60
yet, but from things I've seen, it looks like 2.60 is already obsolete
(spammers have already trained themselves to it).  When will the base
rule system go back to being a strong set of its own that doesn't
depend upon you using bayes (since that's not really an option here)?

It looks to me like 2.55 is still rather weak, but that's because
MailScanner seems to have switched to defaulting to use_bayes, where
the last time I checked it defaulted to not using it.  I've just turned
it back off, to see if that helps my scores.

More information about the MailScanner mailing list