Mozilla to use Bayesian spam filtering

[Clint Adams]

> Last I read on the spambayes mailing list, they've done an incredible 
> amount of work in creating a very good evaluation function for the spam/ham 
> determination.  I don't think it qualifies as either the Graham or Robinson 
> method.  They may have started with Graham, but they've 
> extended/expanded/modified/changed it a whole bunch.

Should we be evaluating or adopting it?