bayesian spam filtering and IRC

[Andy Smith]

On Sun, Apr 06, 2003 at 11:55:16PM +0200, Piotr KUCHARSKI wrote:
> On Fri, Apr 04, 2003 at 09:23:52AM +0100, Andy Smith wrote:
> > I'm familiar with the very basic concept of Bayesian filtering, but
> > have little practical experience with it.  I was wondering
> > potentially how useful it could be when applied to IRC, i.e. built
> > into an IRC server?
> 
> Not worth doing. :)

OK, but can you (or anyone else) explain why?

> > Assume the IRC server keeps the last few hundred messages sent by
> > any given user, and when that user quits from the network normally
> > their messages are considered to be examples of good IRC behaviour.
> 
> Spammers would then quit before getting killed by ircop.

I don't really see the problem here.  If they are quitting sooner
then they are spamming less.  If your point is that quitting would
"reset the score" then I'm sure that could be worked around, what I
said was only a rough example.