I'm seeing different SpamAssassin spam scores being reported by MIMEDefang and MailScanner on the same message.
NO. They are NOT the same message. MailScanner is seeing the message *after* mimedefang has modified it. This changes the message dramaticaly if it's been encapsulated, and causes none of the header rules to match.
You'll only get the same results by feeding both the SAME message without any modifications, and without any bayes learning going on in the meantime.
How can this be? I would rather use MailScanner + SpamAssassin than MIMEDefang.
Double scanning causes this problem. Pick one and only one way to call SA. Stop screwing around with two scanners at once. SA doesn't work that way.
Calling it twice will always screw up the scores for the second scanner, unless your first scanner only adds headers, and even that can change things if you're using bayes and don't have a bayes_ignore_header statement for it.
If you really want to test two at once, make sure mimedefang does not modify the message other than to add a header, and add a bayes_ignore_header for it to your local.cf so it doesn't bias the bayes scores.
