[Spambayes] SpamBayes as a gatway solution

[Peter Beckman]

On Wed, 27 Aug 2003, ROTTENBERG,HAL (HP-USA,ex1) wrote:

> I'm interested in this concept as well.  Perhaps a more limited form of
> bayesian filtering could be used, one that has much lower threshholds and
> perhaps using a fixed spam corpus or a very customized list of tokens.  A
> single ham marked as spam would be unacceptable if you are deploying
> something at the organization level.

 What about a modified SpamBayes, where each incoming email address has/creates
 its own database?  That way, each user would still have their own
 individualized database, but they would all use the same codebase.  The
 web interface would recognize them as their email address, and use the
 corresponding SB DB.  The system would use the users incoming email
 address as the key.  As long as you are using BDB, the DB won't take up
 more space than the spam would.  And for users that don't exist, the mail
 server should reject those pieces of mail before they get to the SB
 filtering/tokenizing/scoring.

 You could even have an aliases file, so if one user gets all 3 email
 addresses delivered to a single person, you could create that alias and SB
 would use only one database for 3 incoming email addresses.

 Then again, my DB is 2.5MB.  Most systems should be able to handle that
 though... the amount of spam not delivered should make up for the disk
 space for individual databases.

 Thoughts?

---------------------------------------------------------------------------
Peter Beckman                                                  Internet Guy
beckman at purplecow.com                             http://www.purplecow.com/
---------------------------------------------------------------------------