[Spambayes] Feature Request

[Michal Vitecek]

Harold Vandeventer wrote:
>I'm confident the lack of white listing within SB is fine.  
>
>The ability to spoof addresses makes the concept of white and black
>lists a waste of time.

 this is very arguable - at least.

>As an example, my own name is being used to send ME email.  Three times
>every month I allegedly send myself email, from servers overseas,
>attempting to sell myself very cheap Microsoft software and watches.
>That mail is also going to some of my co-workers.
>
>If my co-workers were to white list my name, they would continue to
>receive the junk mail. If they blacklist me, they won't get any of our
>business mail.

 of course, everyone can use stupid rules like that and it's then their
 fault if the application does what you described. every solution has
 its pros and cons.

>Only the SB approach of training on the total message content can
>address the problem of spoofing.

 this is true. however have you tried to teach spambayes that some mails
 are ham with a largish ham/spam database? it's a nuisance as spambayes
 must be taught many (and i mean **MANY**) times until the token
 classifications change in favor of the messages. also this is imo
 leading to lower spam probability for certain tokens which is also
 questionable as for the result on overall spam/ham ratio.

 i don't think a staightforward implementation of whitelisting (i.e. if
 a whitelist token is found => message is ham) would do much good but if
 some heuristic were used it could give spambayes very important hint.
 and every additional bit of information is imo a good thing.

        thanks,
-- 
		fuf		(fuf at mageo.cz)