Re: Spam.el, ham-marks

[Ted Zlatanov <tzz@lifelogs.com>]

On Wed, 28 Feb 2007 15:45:42 -0800 Jan Hoffmann <hoffmann@cip.ifi.lmu.de> wrote: 

TZ> this may be obvious, but did you also load and enable the
TZ> gnus-registry?

JH> Oh yeah, sure!  I use the registry independently from spam.el and it
JH> works well.

TZ> What version of Gnus do you use?

JH> I use Gnus v5.11 on GNU Emacs 22.0.50.1 (i386-apple-darwin8) of
JH> 2006-07-08.  Do you think it is a good Idea to start using a version
JH> of the head branch?  If seen that there are many improvements in the
JH> spam.el already.  But apart of that I'm happy with the stable branch.
JH> Maybe I can just install a nwe version of spam.el from the head?  Are
JH> there any revisions of the head that are known to be very stable?

TZ> I'll debug this with you.  You're right, the expected behavior is that
TZ> old articles should not be registered twice.

JH> Thank you.

OK, turn up gnus-verbose to 10 and send me the messages you see when
you exit the group buffer.  If possible, send me your registry file (I
know it may be large) and your Gnus setup.

JH> By the way: It would be nice to have a feature to split the mails
JH> depending on their spam score, i.e. to split mails with a score over
JH> 0.95 to the spam.likely group,  those between 0.75 and 0.95 to
JH> spam.maybe, ...

That's possible.  We already have a function to get the spam score.

JH> Another thing that would be cool would be to have a customization of
JH> gnus-spam-process-newsgroups in terms of the spam score.  I would like
JH> to set for example (ham spam-use-bogofilter "0.4" "2") in ham groups.
JH> That shall mean to learn only from messages that have a high spam
JH> score (let's say over 0.4) to improve the wordlist.db.  The "2" shall
JH> mean to learn it 2 times.  I've read that could be helpful and some
JH> people like it.

Also possible.  Great ideas.  It will be difficult to implement them
with the current customization interface, though.

Ted