Optimizing scoring Re: Exclude from RCVD_IN_DNSWL_MED

[da...@chaosreigns.com]

On 09/17, Kris Deugau wrote:
> As an ISP mail admin, I **CANNOT** afford to block legitimate mail
> from any source, and if I see a report that a legitimate mail was
> blocked by any local rules or DNSBL data, I change the local rule or
> delete the offending local DNSBL entry ASAP.

Some times I envy the data available to those of you with users.  If you
can get 100,000 spams, and 100,000 non-spams together, you could run the
SA results through the re-scorer used to generate sa-updates, and have
scores fully optimized for your own users.  

And then you could give that data to the SA project to make it more
accurate for everybody else.  

I still feel like there's some good opportunity along these lines for
shared bayes.  

-- 
"Democracy is the theory that the common people know what they want,
and deserve to get it good and hard." - H. L. Mencken
http://www.ChaosReigns.com