Jim Maul wrote:
This will not work as all it does is assign -100 points to the message. This could cause an autolearn=ham on every message, even the spam that people post to the list!
Sorry Jim, but that's 100% pure fallacy.. The autolearner explicitly ignores whitelist and blacklist scores.
No need to be sorry, it was my mistake, and im actually quite glad that it does indeed work this way.
To quote the manpage for Mail::SpamAssassin::Conf while describing "***bayes_auto_learn"*:
Note that certain tests are ignored when determining whether a message should be trained upon:
- rules with tflags set to 'learn' (the Bayesian rules)
- rules with tflags set to 'userconf' (user white/black-listing rules, etc)
- rules with tflags set to 'noautolearn'
Therefore, whitelist_from_rcvd will have no impact whatsoever on the autolearning of a given message. Period.
Ok, so if the autolearner ignores the -100 from the whitelist_from_rcvd and uses the score without the -100 to determine whether or not it should be autolearned, what is the point of adding the whitelist_from_rcvd entry at all? I understand that it will pretty much prevent the list messages from ever being marked as spam, but this i am not concerned about.....
However, there is still the danger of a message that contains a few bits of spam gets autolearned as ham because the rest of the post is very ham-like. The only way to avoid this risk is to not run posts to this list through spamassassin at all. (A generally better option, but one that is not supported by all kinds of setups, especially simple milters)
This is exactly what i am trying to prevent. I really couldnt care less if the list messages get marked as spam. What i DONT want to happen is list messages to get autolearned as ham. Am i correct in saying that adding the whitelist_from_rcvd will not prevent this from happening?
Thanks,
-Jim
