Re: NNTP spam filtering

[Ted Zlatanov]

On Tue, 02 Nov 2004, dmaze@mit.edu wrote:

> Ted Zlatanov <tzz@lifelogs.com> writes:
> 
>> When you say the thread makes you confused, please tell me
>> specifically what you want to do and what's not working.  The
>> spam-autodetect stuff is in the manual, I'm sure of it.  You should
>> use the latest CVS.
> 
> If you disbelieve in Customize, it's really hard to find correct
> values for some of the things, and more confusing that there are at
> least three ways to set the things that are group parameters ("most of
> them", and they have a dedicated variable, or you can use
> gnus-parameters, or you can use e.g. 'G p' from the group buffer on a
> per-group basis).

I don't think I can do much about that, unfortunately.  Sometimes
gnus-parameters and group/topic parameters complement each other
nicely, though, so I can't honestly say I'd choose just one or two
configuration styles over the other.

What would be REALLY nice, I think, is a way to run through ALL the
groups and print out the parameters that Gnus knows about them.  It
would let users see their setup in action, and it's a great debugging
tool.

> For autodetect, does this look right?  ("nndsc" is a private backend
> that's not really useful outside of MIT, but it's also read-only in
> the same way nntp is.)
> 
> (setq gnus-spam-autodetect '(("nntp.*" . t) ("nndsc:.*" . t))
>       gnus-spam-autodetect-methods '((".*" . (default)))
>       spam-use-stat t)

That looks right, but let me explain to the other readers:

1) auto-detect spam in all groups that begin with "nntp" or "nndsc:".
   This means that entering that group will always check ALL articles
   to see if they are spam.  You can use the spam-process,
   spam-process-destination, ham-process-destination, and ham-process
   parameters to decide what will happen to that spam when you exit
   the group.  The "move" operation that happens for spam in mail
   groups becomes a "copy" operation with read-only backends such as
   NNTP, but otherwise the spam/ham processing is equivalent.

2) use spam-stat.el for spam detection

3) in all groups, the spam auto-detect methods should use the default
   behavior, which acts as if you called (spam-split) on its own.
   That means that spam-split will use whatever backend it finds to be
   enabled; in this particular case it will use spam-use-stat because
   that's enabled.  The equivalent spam-autodetect method, specified
   explicitly, would be

'((".*" . (spam-use-stat)))

Ted