Raised by Jukka Manner on the mailing list (and my reply):

> - It is sometimes difficult to follow, what is meant by "flow" in the
>   document. Mostly it seems to be used for data flows, but 
> sometimes it
>   seems it is used in signaling flows, or left otherwise 
> unspecified. A
>   couple examples: p.20 6.A "for that flow" (what flow?), 
> p.22 "Section
>   4.2 describes the per-flow and" (what flow?).
> 
>   SUGGESTION: add "data" in front of "flow" where the text 
> really talks
>   about the data flow being signaled for.

this is a tricky one (terminologically): I think the word 'flow'
is always referring to the thing being signalled for; the problem
is that this is only a data flow (in the traditional sense) for 
the path coupled MRM. For example, in signalling for the NATFW REA 
case there is clearly per-FOO state, but FOO is not really a data
flow (since no flow yet exists).

I have added an issue to the tracker for this.