Jo,
I think, this is not the answer to the exact question.
If I understood the question correctly, Ronan asked whether sa-learn
would ignore the headers that were inserted in mails that have been
SCANNED, but probably not yet LEARNED.
If I am right, then the answer might be found in the documentation at:
http://wiki.apache.org/spamassassin/BayesInSpamAssassin
"It's OK to feed emails with Spamassassin markup into the sa-learn
command -- sa-learn will ignore any standard Spamassassin headers,
and if the original email has been encapsulated into an attachment it
will decapsulate the email. In other words sa-learn will undo any
changes which Spamassassin has done before learning the spam/ham
character of the email."
HTH
Christoph Reichenberger
On 12.06.2006, at 14:15, Jo wrote:
Ronan McGlue wrote:
just to check...
I currently use sa-learn by getting a cross-section of my userbase
to copy ALL their spam into a shared imap folder. This bypasses
any extra headers being added if they were to forward etc.
Some of the messages, (the majority) will have already been
scanned by bayes and have a score assigned. Even though there are
now SA headers in the mail does this affect the baysian learner,
or is it smart enough to remove / ignore any SA tags it finds!?
Thanks
Ronan
The Bayesian learner will ignore messages it has already learned
before.
Jo
