[Mimedefang] SpamAssassin 2.40 experiences?

[Michael Sofka]

On Wednesday 04 September 2002 01:26 pm, you wrote:
>
> Has anyone heard any more about the Bayesian analysis that was all the
> rage a few weeks ago?

I started to experiment with it last week, and the results were
encouraging.  My plan is to use MIMEDefang to run bogofilter
(http://www.tuxedo.org/~esr/bogofilter/) on messages.  To allow
for per-user adjustments, mimedefang-filter would look for
mail to spam@<server> and nospam@<server>, and run the messages
through bogofilter adjusting the spam and nospam lists.  The
sender of the message would be used for personalized weights.
(This is on a test machine.)

Alas, personal responsibilities put a hold on the project, for
awhile (I needed to take a few days off during the gab between
the end of daycare, and the first day of kindergarden. :-)  But,
I hope to get back to it soon, since the approach looks promising.


BTW, we are increasingly recomending SpamAssassin Pro to users
who are really really annoyed by spam.  The advantage is it runs
on their desk top, and (if they use Outlook), they can quickly
adjust the white and black lists.  Those who have used it are
very happy with the results.  But, we have no immediate plans
to run server-based spam filters.  The issues are just too
complicated and the false alarm rates too high.

Mike