Display configuration / Improve spam filtering

[Eddy Ilg <ed...@fericom.net>]

Dear spamassassin users,


we are using spamassassin on a site-wide basis. It is invoked via a 
script from postfix and is simply called by "spamassassin -e -x".

Suprisingly, when executing it by a user named filter and executing it 
by root, different tests are executed (See below). Why could that be? 
Especially the HTML_FONT_BIG should be in both outputs. The user_prefs 
file is empty for both users.

Is there a way to make spamassassin report the currently active 
configuration settings and loaded modules (also available perl modules), 
to check the configuration?

We are still getting a lot of spam. We already use AWLs, blacklists and 
Bayes and we have also read the "SpamAssassin" book as well as the FAQs 
and HOWTOs, but we have still not found any measures to improve the 
settings. E.g. the mail that was filterd with the tests given below is 
detected as spam  by a simple outlook filter, but spamassassin doesn't 
detect it.
Any hints what we could do? You can find the config below. We are 
running sa-update regularly.

We don't want to use sa-learn, as the mass of emails is too large. Can't 
bayes improvements be streamed from a central location?


With best regards,


Eddy Ilg



--------------
executed as root:
-0.0 NO_RELAYS              Informational: message was not relayed via SMTP
  0.0 HTML_MESSAGE           BODY: HTML included in message
  0.3 HTML_FONT_BIG          BODY: HTML tag for a big font size
-0.0 NO_RECEIVED            Informational: message has no Received headers


executed with user filter:
-0.0 NO_RELAYS              Informational: message was not relayed via SMTP
  0.0 HTML_MESSAGE           BODY: HTML included in message
  0.0 BAYES_50               BODY: Bayesian spam probability is 40 to 60%
                             [score: 0.4302]
-0.0 NO_RECEIVED            Informational: message has no Received headers

local.cf:

required_score           7.0
report_safe             0
use_bayes               1
bayes_auto_learn              1
skip_rbl_checks         0
use_razor2              1
use_dcc                 1
use_pyzor               1
ok_languages            all
ok_locales              all