Granularity (Was: New list of discretionary items)

DM>I think that yes/no is too much to hope for. It would be very nice
DM>if we could get the answers down to a range of keywords.
DM>Conceptually, this would be like "What languages do you support for
DM>sorting?".

GKH>However, can we not scope our *first* version of the test suite to
GKH>be as  simple as possible, perhaps even deferring cultural sort
GKH>schemes?

We could start with a yes/no question: Do you sort "en-US" alphabet?

However, I am seeking consensus on the point that we will not be able
to represent all aspects of developer discretion with just yes/no or
similar binary (ignore/error) values. Let's try a different item:
the output of namespace declarations. The developer may choose to
(1) output a declaration on every node where that namespace is in
scope (gasp), (2) output a declaration in a minimalist way on the
"highest" node for which it's in scope, or (3) output a declaration
whenever it is stimulated to do so, possibly resulting in duplicate
declarations on a node when its ancestor already has the same
declaration. Other less deterministic outputs would also conform.
I think that we could represent this as one question with multiple
keywords and capture the behavior better than two or more binary
questions.

In my next message: tagging the specs.
.................David Marston