cyrus and Spamassassin bayes rules

[Moritz Both]

We used to use bogofilter with the old pop3 system and with cyrus imapd 
now we don't have any spam filter so far except the one built into mozilla.

A key question to me is if it is a good idea to share a single spam/ham 
keyword database or to have one for every user. This is a bit OT - sorry 
  but I think it can help finding the ''correct'' aproach with cyrus.

- Does it basically work to share the database? If users have different 
opinions about what is spam and what is not,  wouldn't they work against 
each other, feeding messages to the spam filter trying to convince it to 
classify messages ''correctly''. This could for example be the case for 
delivery error messages caused by worms.

- I would assume it is not the responsibility of the system 
administrator to train the global spam filter.

So we'd prefer individual user spam filter databases. This also means 
individual places in the file system for the databases - assuming you 
don't have unix users for all imap users on the machine.

Maybe we will put the spam filter databases into 
/var/lib/bogofilter/user or so, and let it read individual user cyrus 
"learn-spam" and "learn-ham" imap folders. What I like from this thread 
is the idea of using imap to read the training messages and feed them to 
the filter. This also takes care of the question which users exist at 
all - we can read this info from the mailbox list using imap if logged 
in as an admin user.

This is as far as I have got...

Moritz



Ghislain ADNET schrieb:
>   The spam is still something that plagues us and so i installed a 
 > [...]
---
Cyrus Home Page: http://asg.web.cmu.edu/cyrus
Cyrus Wiki/FAQ: http://cyruswiki.andrew.cmu.edu
List Archives/Info: http://asg.web.cmu.edu/cyrus/mailing-list.html