user AND global databases?

[David Relson]

At 09:09 PM 3/14/03, elijah wrote:

>On Thu, 13 Mar 2003, Todd Underwood wrote:
>
> > ...
> > most people implementing bogofilter for their client base are either
> > going to do a sitewide filtering database (poor scaling, less accuracy,
> > not as cool), or per-user databases.
> > ...
>
>I have been wondering about that. Has anyone had success doing both?
>
>I was hoping to do this:
>
>- incoming messages are bogofiltered using a global database
>   the cutoff is set loose, to create more unsure results.
>
>- if the result is unsure, then run bogofilter using
>   a user's private database. the cutoff is set tighter.
>
>- if the result is yes to either one, mark message as spam.
>
>I read in the todo that there is some plan to support multiple databases
>in one pass. How far away is this? Should I just not bother and wait for
>this feature? Would this feature be useful for combining global and user
>databases?
>
>Both bogofilter passes could be -u. Alternately, whenever a user's private
>database is update, also update the global.
>
>-elijah
>
>p

Elijah,

An intriguing two-step approach.  It's certainly very interesting and might 
work well.

Bogofilter has the beginnings of code for working with multiple 
wordlists.  The mathematical foundation needs to be established before it's 
ready for use.  I hope to have more info soon and will write it up for 
discussion at that time.

David