It needs to be made clear how applications should interact with OpenFlow plugin - especially for flow, meter, groups programming. In particular things needed:

1) How to program flows which are dependent on a particular meter or a group.
2) How does an application find that the flow it sent was indeed programmed on to the switch.
3) The different methods of programming the flows:
a) Via data store (GBP, etc.) uses this mechanism. Pluses include preprogramming of flows, etc. Minuses - performance may be an issue, getting information on whether the flow programming succeeded is tricky.
Also add details about RESTCONF flow programming (what happens when the REST API is called to the flow is sent on to the switch.
b) Using RPCs (L2 switch uses this mechanism). Pluses include performance, getting information on whether the flow programming succeeded; Minuses: preprogramming, etc.

4) Operational vs Config data store - how & when does the flow get to operational and/or config datastores.

5) Turning stats collection for a flow on/off.

Please include any other things which I have missed but should be documented.