USER_IN_WHITELIST and autolearn

[Alexis Manning <al...@googlemail.com>]

It seems that if USER_IN_WHITELIST is triggered then the message won't be
auto-learned.

 

X-Spam-Status: No, score=-100.0 required=7.5tests=BAYES_50=0.001,

            HTML_MESSAGE=0.001,USER_IN_WHITELIST=-100 autolearn=no
version=3.1.7

 

I have a fair number of people in my whitelist and I would like to feed that
ham to the Bayes database: is there any option which will allow their
messages to be auto-learned too, or do I have to do it manually?

 

Thanks,

 

-- A.

Re: USER_IN_WHITELIST and autolearn

[Alexis Manning <al...@googlemail.com>]

mkettler_sa@verizon.net says...
> Alexis Manning wrote:
> > It seems that if USER_IN_WHITELIST is triggered then the message won?t
> > be auto-learned.
> >
> That is incorrect, however USER_IN_WHITELIST does not count toward any
> autolearning decisions.
[...]
> As far as the autolearner is concerned, this message scored 0.001.
> (BAYES_50 doesn't count either, to avoid bayes self-feeding.).

Matt, thanks very much for the comprehensive reply.  Off to read about 
AutoLearnThreshold now :)

Cheers,

-- A.

Re: USER_IN_WHITELIST and autolearn

[Matt Kettler <mk...@verizon.net>]

Alexis Manning wrote:
>
> It seems that if USER_IN_WHITELIST is triggered then the message won’t
> be auto-learned.
>
That is incorrect, however USER_IN_WHITELIST does not count toward any
autolearning decisions. The primary reason being that whitelists are
often misconfigured the first time around (ie: whitelist_from
*@mydomain.com). If the whitelists counted towards autolearning, people
making whitelisting mistakes would have a lot of cleanup work to do
fixing the bayes DB as well.

As far as the autolearner is concerned, this message scored 0.001.
(BAYES_50 doesn't count either, to avoid bayes self-feeding.).
>
>  
>
> X-Spam-Status: No, score=-100.0 required=7.5tests=BAYES_50=0.001,
>
>             HTML_MESSAGE=0.001,USER_IN_WHITELIST=-100 autolearn=no
> version=3.1.7
>
>  
>
> I have a fair number of people in my whitelist and I would like to
> feed that ham to the Bayes database: is there any option which will
> allow their messages to be auto-learned too, or do I have to do it
> manually?
>
They will be autolearned, but only on their own merits. The whitelisting
won't force autolearning as nonspam.

You might want to read the autolearner documentation to see how the
autolearner calculates score. It changes a few things and the score it
uses can be quite different than the final message score.

http://spamassassin.apache.org/full/3.1.x/doc/Mail_SpamAssassin_Plugin_AutoLearnThreshold.html

In this case, all the whitelist rules have the tflag "userconf".


>  
>
> Thanks,
>
>  
>
> -- A.
>