I share Bernard's concern about the kind of graph we end up with,
because an ISO standard is for everyone, not just for current
implementers. It would be good to be able to use established
mathematical techniques to gain assurance that the RM graph has nice
rather than nasty properties for future implementers.
I deleted "well formed" on the basis of my earlier comments on parid0484. I don't think it is
necessary to invent neologisms to avoid implying data structures. "Relationship between" serves just
as well and does not imply any data structure. Less jarring to read.
I think loop or "self-loop" is the term we want from graph theory to say that a node may serve as only
one end point of an arc. Seems easier to just say that, although I am not sure my prose is any more
readable than the original. More explicit perhaps.