few questions

[Matthias Andree]

Pierre-Philipp Braun schrieb:

> I'm using bmf, another bayesian mail filter for years (bmf.sf.net).
> But now I need a milter interface, and bogofilter has one: bogom 
> (www.usebox.net/jjm/bogom/).  A few questions:
> 
> - Can bogofilter update its db itself (self-learning) ?  In example, when it
>    figures out the message is a spam, does it add message's words to the bad
>    words db ?

That would be bogofilter's -u option.

> - I need a system wide wordlist for inclusion in a mail server (as said, as
>    milter even).  Are there any up-to-date wordlists available on the internet?
>    Or at least some minimalistic wordlists to have a fresh start ?  I doubt
>    using my personal (user) messages to build the db is a good idea.

I don't know of one, but perhaps somebody else does.

Otherwise, setting up a system where all users can correct
misclassifications, registering a few hammish messages and a few spamming
should suffice to give a start. There are several scripts around to help
training on attachments of mail submissions, for instance, so that you can
set up spam or ham registration addresses on your central mail hup where
users just forward misclassified messages as attachment.

HTH,

-- 
Matthias Andree