On Tue, 20 Jun 2006, Matt Hampton wrote:
I have had a quick check of the archives but can't see any relavent
threads.
I would like to be able to find out what effect a token would have
on Bayes scoring.
For example - I want to be able to find out whether a header that
I am inserting before SpamAssassin sees the message is having a
positive or negative affect and if so by what magnitude.
I am adding the AS Number and CIDR by a milter. I want to track
which AS Numbers are producing the most spam.
As I understand it, only hashes (checksums) of keywords are
added to the bayes keyword database. Therefore, it is not
possible to retrieve a list of keywords and their values.
The closest thing you could do is log all the keywords that
you add to the messages separately. Then you could use
the same algorithm for generating hashes that SA does and
look those up in the bayes database and correlate that way.
I don't know what the algorithm is, but it should be fairly
easy to determine by digging through the SA source code.
- Logan
