spambayes rocks

[Skip Montanaro]

    Milan> I've only clarified that solutions like spambayes help not only
    Milan> us, but also spammers.

For systems like SpamAssassin, where the database of spam/ham information is
stored centrally, I agree.  Spammers "only" have to tweak their messages to
thwart the contents of a single database.  With Spambayes and other similar
systems, spammers have no access to the spam/ham information, as it's stored
person-by-person.  If they are able to craft messages which are good enough
to sneak by, I suspect it's quite possible they are selling something I
might actually be interested in or know a hell of a lot about me.

    Milan> In a longer term, we have to find much more efficient combat
    Milan> methods against spammers.

I assume by "efficient" you mean closer to 100% accuracy.  At this point
Spambayes is there for me.  There's not much room left.  When a new spam (or
virus) comes out, it only takes me training on a few of them to squash them.
With SpamAssassin, I had to either manually develop my own rule or wait for
the next SA release.

None of this is meant as any sort of indictment of SA, by the way.  I used
it quite happily for a long time.  It was the first tool I encountered which
had any real success against spam at all.

Skip