Re: Fight spam with BBDB database?

[David Z Maze]

Richard Wellum <rwellum@cisco.com> writes:

> It's just an idea, wanted to see if it was possible. Could I filter all
> email from users not in my BBDB database into a SPAM folder? That way I
> could run through the SPAM folder periodically, if there is a legitimate
> email in there I simply add to my BBDB database, and the next time it
> will go into my Inbox?

That's not a great way to hunt for spam, in my experience, especially
if you include "viruses" in "spam".  (In particular, if you subscribe
to open mailing lists, you'll get spam addressed to the mailing list
and won't be able to tell it from legitimate mail to the list without
extra filtering.)  But this strategy is still somewhat close to what I
do, which looks sort of like this:

(defvar dzm-bbdb-known-group "mail.known")

(defun dzm-bbdb-split-to-known ()
  (cons '&
        (apply 'nconc
               (mapcan
                (lambda (rec)
                  (let ((nets (bbdb-record-net rec)))
                    (cons
                     (mapcar
                      (lambda (net)
                        (list 'any (regexp-quote net)
                              dzm-bbdb-known-group))
                      nets))))
                (bbdb-records)))))
;; Gee, I must have written that around when I was taking 6.001.

(setq nnmail-split-fancy
 `(| ;; note backquote
   (: spam-split) ;; using spam.el
   (&
    (to "debian-\\(\\w*\\)@lists\\.debian\\.org" "debian.\\1")
    (to "ding@gnus\\.org" "gnus.ding")
    ;; etc.
    )
   ,(dzm-bbdb-split-to-known)
   (any "dmaze@.*mit\\.edu" "mail.personal")
   ;; anything else:
   "mail.impersonal"))

So, the sorting rules sort to the first of:

(1) Spam groups as defined by spam-split;
(2) Mailing lists, cross-posting appropriately;
(3) People I know into mail.known;
(4) Mail addressed to me into mail.personal;
(5) Anything else into mail.impersonal.

(The actual implementation of (3) uses (bbdb-record-getprop rec
'mail-group) to get the value of the mail-group field of each record,
tests if that's set, and if so emits splits pointing there as opposed
to elsewhere.  So, one group per person, with crossposting.)

-- 
David Maze             dmaze@mit.edu          http://www.mit.edu/~dmaze/
"Theoretical politics is interesting.  Politicking should be illegal."
        -- Abra Mitchell