One major problem I found with bayes_ignore_header is with regard to
lotus notes.  Most of my users are notes users.  My mail gateway is unix
(of course) running SA.  When users receive mail that is incorrectly
marked they used to send it to a dummy user on the gateway to be learned
by sa-learn.  However, I found that notes added lots of headers that
bayes_ignore_header did not remove.  On top of that notes
changed/modified some of the headers.  And if a user did not send the
mail correctly then no headers were send. (Notes strips headers
assigning them to variables thinking no one will ever want to see them.)

In the end I talked to our notes developer and found he could write a
simple agent that would take the current message grab the original
headers and drop that off to a dummy user.  The dummy user is really a
program that looks for markers created by the agent telling me where the
true start and end of the message is and spits out the original message
with the original headers.  I then have cronjob that runs the spam/ham
through sa-learn.

Louis



On Thu, 2003-06-12 at 11:57, Duncan Findlay wrote:
> On Thu, Jun 12, 2003 at 01:37:01PM +0100, Adam T. Bowen wrote:
> > Hi,
> > 
> >     I have created a mail box for users to bounce spam, that has
> > slipped through SpamAssassins net, to.  I want to manually sort through
> > this box and feed the proper spam to the Bayes classifier.  The problem is
> > that when the mail is bounced (and forwarded because sometimes people just
> > don't listen) the headers change.  I am worried about some of the headers
> > that are added being used as a 'short-cut' by the Bayesian classifier.  
> > So the question is, what headers are included or excluded from the
> > Bayesian process?  I would imagine having a spam mailbox for people to
> > bounce to is quite a common way of collecting food, for the Bayesian
> > monster.  What precautions are taken by others?  A chair and whip?
> 
> You could use the bayes_ignore_header option to ignore headers you
> don't want to be considered by bayes. There are a bunch ignored by
> default too.
-- 
-
--
---
¤¤º°`°º¤ø,¸¸,ø¤º°`°º¤øø¤º°`°º¤ø,¸¸,ø¤º°`°º¤øø¤º°`°º¤
¤°`°Lightbridge, Inc
¤°`°67 South Bedford St.
¤°`°Burlington MA 01802
¤°`°781.359.4795 mailto:[EMAIL PROTECTED]
¤°`°http://www.lightbridge.com
¤¤º°`°º¤ø,¸¸,ø¤º°`°º¤øø¤º°`°º¤ø,¸¸,ø¤º°`°º¤øø¤º°`°º¤



-------------------------------------------------------
This SF.NET email is sponsored by: eBay
Great deals on office technology -- on eBay now! Click here:
http://adfarm.mediaplex.com/ad/ck/711-11697-6916-5
_______________________________________________
Spamassassin-talk mailing list
[EMAIL PROTECTED]
https://lists.sourceforge.net/lists/listinfo/spamassassin-talk

Reply via email to