SpamAssassin: Server Wide Setting What does it do? Training?

[rampy]

FWIW: I've been in the plesk admin/client manual, KB/FAQ, and done some googling/forum searching but it still isn't clear to me... so bear with the newbish question:

What exactly does the SA server wide setting do for you? Did plesk load SA up with some rules?

If you set the server SA settings to server wide how do you "train it" (I see the training facilities/options for personal per mailbox SA if configured that way) but it's not clear what to do about training the bayesian filter for the server wide setting..

Side question: What does a "hit" mean exactly? Is that in reference to SA scoring? is higher more or less tolerant of spam?

Furthermore why isn't this in the Plesk UI help or in any of the documentation? It's a pretty basic/obvious question.

My main question is this: Is there a way to train the server wide SA setting of spam and ham? When someone is using the server wide setting how do they tell the server about false positives or fix scoring issues or identify spam SA missed? (and yes I see SA training if personal settings is enabled)

thanks in advance for any light or insight the community can offer. (oh almost forgot on plesk 7.5.4 w/FC2)

rampy

 

[phoenixisp]

Serverwide settings are just that. In my case, most users don't bother to log in and change their personal settings or train SA. So basic rules are set for evryone that has turned spamassassin on. These rules are spelled out in the local.cf file. For instance, you may want to set a score of 6 for spam to be tagged serverwide, yet as a user you may wish scores of 5 tagged, or even deleted.

However, with Plesk's version of Spamassassin serverwide training is not allowed for individual users, only for their particular mailbox. This is very logical, if user A marks emails from ACME Cleaners as spam, yet I do regular business with them ....I can't get the mail because they are blocked serverwide.

Spamasassin is capable of serverwide learning using bayes, but you would have to configure that on your own - Plesk doesn't.

 

[rampy]

Thanks for the response. It's appreciated.

I think you nailed it on the head with the example of the user/mailbox who doesn't log in is covered by the serverwide rules/settings, but if they want to add the bayesian piece they'd have to log in and do the training.

I guess I didn't realize SA did more than just bayesian filtering to analyze spam.

thanks again,

rampy

 

[JP Kelly]

where is server wide db?

Where would the bayes db be located for server wide settings?