> The changes made were to make all types into named types, to
> move certain data types into attributes and to make the
> element names unambiguous at the top level (hence adding
> Status to everything).
>
> It does not matter much what the coding stye is, but making
> it consistent across the spec makes a big difference.
I actually prefer the attribute-based version to the element approach
(maybe Rich can comment on why they went with elements, I don't know
myself), but I do think the array vs. tree issue needs to be
specifically raised.
If we go with a sequence of codes, we need to document the semantics.
With nesting, it's pretty clear that you have a general->specific
connotation.
-- Scott
