We will concentrate on application to application messaging for a single user on a single machine
We will have a hub/daemon process to control the messaging
The messaging system will be kept simple: no encryption, no transactions, guaranteed delivery etc
The messaging protocol will be decoupled from the semantics of the messages
Areas where we probably agree
The messaging protocol will be defined independently of the details of any concrete implementation (this does not preclude such details also forming part of the specification)
Applications are free to spawn/include an instance of the hub/daemon, but should behave "nicely" (TBD) if they shut down or one is already running.
We won't attempt any clever hand-over from one hub/daemon instance to another.
Applications will be located dynamically (through the hub?), rather than from a remote or local registry
The mechanism used to bootstrap the connection to the hub/daemon will be simple
Areas still to be discussed/agreed
A representative list of use cases (see follow up)
Precisely how we bootstrap the connection to the Hub/daemon
The messaging protocol
Which "wire" protocol(s) to use
Pretty much everything to do with the messages themselves