On Nov 10, 2006, at 10:48 AM, Patrick Durusau wrote:
> A further wrinkle: Note that text:id is defined as string not ID. I
> have not checked on the others.
Even if the datatype is defined as an ID, though, that has no formal
meaning in RELAX NG. A validator will not, for example, check that
the values of the attributes are unique. My understanding is they
will if they understand xml:id.
> I will have to re-read the spec but to change the interpretation of
> text:id to be ID may pose backward compatibility problems.
May I suggest, then, that we put this (the meaning of the existing id
attributes, and the possiblity to move to xml:id in the future) up
for discussion soon?
We can't support Rob's "extrinsic metadata" use case (metadata
statements that reference document fragments; rather than vice versa)
without resolution here.
Bruce
