Is there any way to make SA -not- auto-learn when the "USER_IN_ALL_SPAM_TO" rule applies to the message?
I have some customers who I have configured SA to NOT filter for. These customers want to receive SPAM because they're afraid they'll miss important e-mails if they don't. (Not my choice, just work with me here.) However, I know they're getting SPAM, and since they're set up as all_spam_to in the configuration file, they're all coming up with negative scores, and therefore a lot of these messages are being user to autolearn. I think this is greatly throwing off my bayenesian filters.
The score of ALL_SPAM_TO should not be a part of the autolearner's decisions, thus this should not be a problem.
Other than expectations do you have any reason to believe the autolearner is misfiring?
