Advice on Using sa-learn

[Errol Neal]

At 11:31 AM 10/16/2003, you wrote:
>At 09:54 AM 10/16/2003, you wrote:
>>What about Eudora's redirect? Will that work also?
>
>First, no.. Eudora's redirect builds a new set of headers. Not gonna work.
>You need the original message headers, since SA's bayes engine learns
>header tokens as well as body text tokens.
>
>Also see my previous post in this thread. Eudora destroys part of the body
>of any multipart email when it receives it.
>
>That portion is gone and is unrecoverable, so it doesn't matter how you
>redirect, forward, or whatever, important parts of the mail (from a bayes
>perspective) are just plain gone. Most spam is multipart/alternative type
>mail, and won't be usable for learning.


Yeah. I saw your post just after I let this e-mail go. Your final statement.

>>Most spam is multipart/alternative type
>>mail, and won't be usable for learning.

Please explain if you don't mind. Do you mean AFTER it's been retrieved by a
MUA or are you speaking about its characteristics in general?


Errol