[RFC] 'bzr-email' using revision_id for threading

[Stephen J. Turnbull]

John Arbash Meinel writes:

 > There are certainly things I could do to make them more unique. But
 > the *point* was to have them such that 2 clients could create
 > threads for eachother. So that if Robert commits a patch that I
 > then build upon, you can follow that thread.

Which means what?  Who do you propose be allowed to generate the
particular *email* message that starts the thread?  Do you want people
to be able to somehow "join" the thread simply by specifying an
appropriate Message-ID?  ... You can see how fast this gets messy.

Email message IDs must be unique; this means that a client may throw
away a message containing a seen ID, or throw an error, or anything --
behavior is undefined.  If different clients are potentially referring
to a given revision, but their commentary is significant, then the
revision reference can't be used as a message ID except in
combination with uniquifying information.

 > And you could see which pqm revision the work was done against, etc.

Why not just use X-Bzr-whatever headers?