On Mon, Dec 29, 2003 at 03:38:22AM +0100, Carlos E. R. wrote:
The Saturday 2003-12-27 at 21:19 +0100, Jon Clausen wrote:
And since anything the company server is going to handle is company mail, the criteria for spammyness is a (well, will be anyway) company policy.
Mmmm... not if you allow users to directly update the Bayesian filter database. If you want to keep the company policy, then spam should be forwarded to a person responsible of checking them and feeding the result to sa-learn. Otherwise, you might end by someone sending an email to be considered spam, and another user sending the same email to the list of non spam - the filters would get mad at you :-)
Now *that's* a point. So the question then becomes how to avoid that situation... My first impression would be to set up some kind of 'mutex-logic' to make sure no identical mails exist in the two folders... I don't know. In the end it might actually be simpler (and less cpu-intensive) to maintain user-specific filters... The thing is it's a small company, and I doubt this task of 'spam-screening' would be welcome with *any* of the guys...
<snip>
Who said size doesn't matter? ;D
X-)
I've never said file size is not important, contrary to what many developers and other people think :-)
They think you said that? Shame on them !-) /Jon -- Whatever rocks your boat!