Lists Home |
Date Index |
Elliotte Harold wrote:
> Not really. It varies from one binary XML proposal to the next, but many
> subset XML by, for instance, removing the document type declaration and
> DTDs. Others superset XML by allowing the presence of arbitrary binary
> data including characters like null that are simply not legal in XML.
Yes, that's where properties like "Roundtrippable" are important. Note
that just because some binary XML proposals have made mistakes doesn't
mean that all of them have to. In fact, it's precisely because there are
several years' experience making mistakes that we can probably get on
the same page to write a standard now.
> Reuse of common document models, query languages, schema languages, and
> so forth is good. However, if we have to warp them from their original
> intended purpose of handling XML to fit models that don't have types on
> attributes, or allow null character data, or the other weird things we
> see off in the binary space, then they become more complex and less
> useful. One of the nastiest problems that arises in practice is that
> these tools begin creating and describing document that cannot possibly
> be serialized as well-formed XML. The infoset is a particular offender
> here, and its problems tend to infect downstream specifications. That's
> why I'm much more comfortable working from real XML documents than some
> abstract, infoset formulation.
Yup, same comment.