Filtering and ignorelists

[Tom Anderson]

On Fri, 2004-03-05 at 07:11, Tom Allison wrote:
> But all that aside and still subject to debate...  what is not ok with 
> the current training?

The main thrust of the matter is that many headers are invisible. 
Therefore they are malleable like HTML comments.  We currently strip
HTML comments in bogofilter because of the propensity for abuse by
spammers inserting hammy words in there.  The same effect is achievable
by using "X-mailer: Outlook Express" (or whatever the real line is) and
similar invisible header lines to skew the scoring.  I receive many
unsures due to this effect.

To be consistent with our policy toward HTML comments, invisible headers
should also be ignored.  There is one caveat however in that the
Received header is issued by the delivering mail server and is therefore
a trusted agent which may accurately describe the originating SMTP
server and sender (which would be very valuable).  The arbitrary
X-headers may be a true indicator one way or the other in some cases,
but often they are just red herrings, and should thus be nixed.  Visible
headers are just as important as the body.

That being said however, I'm not sure bogofilter is the right place to
add these non-statistical heuristics to massage the headers.

Tom

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 189 bytes
Desc: This is a digitally signed message part
URL: <http://www.bogofilter.org/pipermail/bogofilter/attachments/20040306/83f6a428/attachment.sig>