You are viewing a plain text version of this content. The canonical link for it is here.
Posted to users@spamassassin.apache.org by Matt Wills <sa...@virtualvermont.com> on 2005/06/26 17:45:09 UTC

What else is necessary for Bayes?

My local.cf is like so:

# Enable the Bayes system
use_bayes             1

# Enable Bayes auto-learning
bayes_auto_learn      1
bayes_path            /home/spamd/
bayes_file_mode       0666

using sa-learn --spam etc., Bayes analyzed 270 messages.

I restarted spamd.

Am I mistaken that testing on some of the analyzed messages should result 
in a Bayes score?

Such doesn't happen. The redirected messages come through with the same 
tests triggered as the originals.

What am I missing?

Matt 



Re: What else is necessary for Bayes?

Posted by Matt Kettler <mk...@comcast.net>.
At 11:45 AM 6/26/2005, Matt Wills wrote:
># Enable Bayes auto-learning
>bayes_auto_learn      1
>bayes_path            /home/spamd/

That bayes_path is probably NOT what you want. Read the docs carefully. 
Despite being named "bayes_path" it's really a path AND a partial filename.

try adding "bayes" to the end, that way you'll get /home/spamd/bayes_toks 
as your bayes db. Currently you'll get /home/spamd/_toks.

>bayes_file_mode       0666

Change that to 777. The file_mode is actually a mask, and is used in the 
creation of temp directories.

  Since it's a limiting mask, and is also limited by SA's request of 666 
for the files, the end result will be 666 and files and won't get the X bit.

However, directories SA will ask for 777 and will get 777.

>using sa-learn --spam etc., Bayes analyzed 270 messages.
>
>I restarted spamd.
>
>Am I mistaken that testing on some of the analyzed messages should result 
>in a Bayes score?

No. You need 200 spam AND 200 ham messages learned before bayes will become 
active.

Learning 270 spams puts you over the spam minimum, but you need to learn 
200 hams as well (or wait for autolearning, and with ham, that's slow)

Be sure to check sa-learn --dump magic to see what your nspam and nham 
training counts are.






Re: What else is necessary for Bayes?

Posted by Kai Schaetzl <ma...@conactive.com>.
Matt Wills wrote on Sun, 26 Jun 2005 11:45:09 -0400:

> What am I missing?

Simply run a "spamassassin -D --lint" to find out ... You probably didn't 
learn it enough *ham*.

Kai

-- 
Kai Schätzl, Berlin, Germany
Get your web at Conactive Internet Services: http://www.conactive.com
IE-Center: http://ie5.de & http://msie.winware.org