[CF-metadata] CF-1.0-beta5: curvilinear bounds "contiguous"attribute

[jonathan.gregory at metoffice.com]

Dear Brian and all

As my other posting says, I don't agree we should reverse our earlier decision
to support (n+1) as well as (n,2). In my view, this would be an unnecessary
complication for applications reading the data. Whether (n+1) is simpler is
a matter of opinion, I'd say. I like (n,2) because both boundaries (i,*) have
the same index i as the coordinate. The extra storage required for (n,2) is
practically unimportant.

It is true that (n+1) implies contiguousness. But this is a simple case anyway.
A 2D case of the form (n+1,m+1) is also easy to check if the vertices are
dimensioned (n,m,4). The difficult cases for contiguousness are not these
rectilinear ones. For cells with complicated boundaries, (n+1) is not
applicable. Therefore I feel reintroducing this convention would complicate
the standard for no great practical advantage.

Best wishes

Jonathan