James Clark wrote:
>
> I've been thinking about a catalog-based approach, and I think that would be
> more powerful than a PI although significantly more complicated. The
> catalog could contain rules like:
Even if it is more complicated, it is a long-term solution. It allows the recipient
not to use the schema of the sender. It allows schema evolution without rewriting
documents and without in-place update of schemas.
I am against PIs or attributes, since they make a long-term solution more difficult.
If we introduce PIs and implementations start to rely on them, it will become harder
to introduce a catalog-based approach.
Cheers,
Makoto
