Re: automatically splitting conversations

[Arnd Kohrs <kohrs@acm.org>]

>>>>> "Kai" == Kai Großjohann <Kai.Grossjohann@CS.Uni-Dortmund.DE> writes:

    Kai> I have devised a patch for Gnus which does the following:
    Kai> whenever a message is written to a group, Gnus memorizes the
    Kai> message id and the group name.  There is a function which looks
    Kai> at the References header of an incoming message and searches
    Kai> the memorized message ids for a group name.  If a group name is
    Kai> found, it is returned.  If non is found, nil is returned.

Excellent, Kai.  And your code reveals that it wasn't as complicated to
code as I thought - by using the id cache heuristic.  However, using
nnmail-cache as a store for where messages are moved or split might
prevent a response to this message - in ten years from now - from not
being split correctly ;-).


    Kai> This does what you want, I think: you just `B m' the message

B m, did not do what was intended, see below.

    Kai> into a group, and all followups (and followups to the
    Kai> followups) go into that group by default.

    Kai> For the patch to work, you have to set nnmail-treat-duplicates
    Kai> to a non-nil value.  And then you have to use the function
    Kai> nnmail-split-fancy-with-parent in your split method.  I use
    Kai> nnmail-split-fancy, as follows:

    Kai> (setq nnmail-treat-duplicates 'delete) ;you might prefer `warn'
    Kai> (setq nnmail-split-methods 'nnmail-split-fancy) (setq
    Kai> nnmail-split-fancy '(| (: nnmail-split-fancy-with-parent)
    Kai> ...other splits go here...))


Besides setting these variable you absolutely need also

(setq nnmail-cache-accepted-message-ids t)

So that moving messages results in an update of the id cache.

Since I use gnus-group-split it is 

(setq nnmail-split-methods 'gnus-group-split)

for me, and then for my default mail group I have the following group
parameter:

(split-spec |
             (: nnmail-split-fancy-with-parent)
             catch-all)

    Kai> I wonder if this is documented.  Hm.  No, apparently not.  This
    Kai> is bad.  Please tell me how it goes when you try this.  I might
    Kai> need your experience for writing the documentation.

This is great functionality - It should definitely be documented.  It
could be labeled as something like /implicit splitting/, since the
user's action ("B m") is observed, and Gnus derives an implicit
splitting rule.


Arnd.
-- 
Arnd Kohrs  -  Institut Eurecom - http://www.eurecom.fr/~kohrs