Re: Endividual user settings for Spam , advice needed

[John Andersen <js...@pen.homeip.net>]

On Wednesday 26 July 2006 00:42, jdow wrote:
> For small installations, where I think SA can be at its most
> devastatingly accurate, per user rules are required. 

> For large installations per user rules become cumbersome time eaters.
> They just don't need to enable it at all, take the accuracy hit, and
> proceed with life doing the best they can.

My take is exactly the opposite.  Small shops are usually focused,
and single purpose, and can live fine with site-wide rules.

Larger companies have many more specialties working together
and the advertising manager would want different rules than
the engineer.  Individual rules are essential.


This is what bayes is for.  The power and accuracy of bayes seems to 
overshadow all the other rules.

The big missing feature in SA is a standardized and reliable way
of training individual bayes databases of users.  If you have a windows
shop with linux mail servers and you don't run imap its almost impossible
to get end users to train their bayes.


-- 
_____________________________________
John Andersen