Aleksander Adamowski <[EMAIL PROTECTED]> wrote:

> Understood, what I wanted to say is that Bayes isn't effective against
> this sort of stuff and currently the other SA mechanisms aren't
> sufficient to catch this spam.

My point was that the extra words have no effect one way or the 
other on the Bayes classification.  If they hadn't been there, 
the message would still have slipped through, so it's not 
appropriate to call the extra words "Bayes poison".  People 
talk about "Bayes poison" a lot, but I have yet to see an 
example that actually affects Bayes.

> This is mainly because HTML.pm can be fooled by dangling attributes.

You've lost me there.  What do "dangling attributes" have to do 
with this case?  HTML_FONTCOLOR_UNKNOWN was triggered, so the 
COLOR attributes were seen.  The problem is they weren't 
recognized as being nearly invisible, so the problem seems to 
be with the HTML_FONT_LOW_CONTRAST test, not with parsing.

-- 
Keith C. Ivey <[EMAIL PROTECTED]>
Washington, DC

Reply via email to