re “colonial” – A name is just a name, and there is no reason to carry the analogy into the actual design, ie “is “thunderbird” really likely to be the source of truth that is propagated to exchange?” As to “I would have thought the reverse”, you are right that at some level “tb and exchange are peers.”. After I started using “native” to name the Mozilla platform Exchange objects, I realized that it was just as valid to view the Mozilla mailnews objects as “native” and Exchange as somehow “foreign”. But they are peers, and who is the native and who is the foreigner depends entirely on your perspective.

By the way, I did a posting on m.d.a.t proposing to call the Mozilla Mailnews layer by a more distinctive name, “Skink”. This conversation is a typical example of where I would love to have a better word for those objects than “Mozilla Mailnews” – bugt I got zero traction on that.

The Zindus addon implements a Mealy Machine:

http://en.wikipedia.org/wiki/Finite_state_machine#Transducers

and experience has shown this to be a very useful design pattern in:
- maintaining consistency
- knowing when consistency is lost
- co-ordinating changes in state across local and remote authorities

re: “colonial” terminology: I can see why you chose it, but it implies a master/slave relation and when push comes to shove (ie when consistency is lost and need to be re-established) is “thunderbird” really likely to be the source of truth that is propagated to exchange? I would have thought the reverse or at least that tb and exchange are peers.