Re: OT: Sorting by references using Sieve

[Ted Zlatanov]

On Thu, 19 Mar 2009 08:52:01 +0100 Tassilo Horn <tassilo@member.fsf.org> wrote: 

TH> What I want is this: If I get a reply by John Doe, it should be filed
TH> into the folder that contains the request I've sent, e.g. the folder
TH> which contains the message with the Message-id found in John's
TH> References header.

TH> Is this doable with Sieve?  I had a quick look at the RFC, but couldn't
TH> find something like that.

I don't think Sieve can do this on its own.  Can it do a #include?

TH> I know that I could use the gnus registry for that, but I'd prefer to
TH> keep the filtering on the server side.  Anyway, it Sieve doesn't work
TH> here, a short sample for a gnus registry/fancy splitting routine would
TH> be appreciated, too.  One key point is that it mustn't interfer with the
TH> server side filtering.  E.g. it should only look at articles which are
TH> delivered to my INBOX, everything else was filtered by the server setup.

Well, the registry could write a Sieve rules file based on its contents.
I don't think it would be easy to integrate them further.  

Another way is to have the registry save its contents to Amazon SimpleDB
or a CouchDB instance, and then the server can generate filters based on
that data.  This way you could actually have both the registry and a
server-side process remembering article references.  Hmm, that sounds
like an interesting project :)

As far as a simple gnus-registry setup, it's pretty easy to do:

(setq gnus-registry-max-entries 2500
      gnus-registry-use-long-group-names t)

(gnus-registry-initialize)

Then use this in your fancy-split:

(: gnus-registry-split-fancy-with-parent)

The fancy-split rules are run against a mailbox, so you need to set it
up to run against INBOX.  That's all.

Ted