[Kolab-devel] [issue786] Add mechanism to teach spamassasin's bayesian filter

[Stephan Buys]

New submission from Stephan Buys <s.buys at codefusion.co.za>:

It would be nice if there was a standard way to teach spamassasin's bayesian          
filter with Kolab.          
          
This could be achieved by running "sa-learn" on a special spam mailbox, etc.         
         
What I have done for our setup is create the following:         
         
1) Add to master.cf.template:         
 spamlearn     unix  -       n       n       -       1       pipe        
user=spamlearn        
     flags= argv=/kolab/bin/sa-learn -L --spam --no-sync --single        
        
hamlearn     unix  -       n       n       -       1       pipe user=spamlearn        
     flags= argv=/kolab/bin/sa-learn -L --ham --no-sync --single        
       
2) Add user spamlearn and make sure passwd looks something like:      
      spamlearn:x:19418:19415::/kolab/var/amavisd:/bin/bash    
    
3) Add to a transport map something like:   
 #Spam table  
/^spamlearn at localhost.invalid$/ spamlearn:localhost  
/^hamlearn at localhost.invalid$/  hamlearn:localhost  
 
The idea is to forward spam that is not tagged correctly to spamlearn or 
hamlearn until the server spamassassin starts to catch it... 
 
(Mine isn't working so well yet ?? :-)

----------
messages: 4844
nosy: stephanb
priority: wish
status: unread
title: Add mechanism to teach spamassasin's bayesian filter
topic: server
________________________________________________
Kolab issue tracker <kolab-issues at intevation.de>
<https://intevation.de/roundup/kolab/issue786>
________________________________________________