[
Lists Home |
Date Index |
Thread Index
]
> The problem is that there can be synthetic infosets that cannot
> possibly be serialized as well-formed XML documents, and other specs
> are being defined in terms of this most general infoset instead of
> the much more restrictive and interoperable case of XML itself.
We should be careful not to assume that data interchange serialization
is the only kind of interop that matters.
I think the situation with XQuery dependence on infoset data model is
much like the SQL dependence on relational data model. In the case of
query models and programming models, the anchoring to a common data
model is extremely important to interop. (And regardless of whether
this was the original intent of Infoset spec or not, the infoset spec is
being treated as a common data model in a new generation of specs)
Serialization format is mostly orthogonal to data model. If infoset is
to XML 1.0 as relational model is to CSV, you can see that it would be
silly to deliberately disallow commas in relational text columns because
"it can't be represented properly in CSV". In fact, it is easy to
imagine an XML 1.0 serialization of any arbitrary infoset that would be
guaranteed to always preserve full fidelity on round-trip. It just so
happens that the *canonical* mapping of infoset to XML 1.0 does not
cover all cases -- and in my opinion that is not such a big deal (no
more a big deal than the situation with relational and CSV today)
|
