Lists Home |
Date Index |
Ok. That's two for syntax-only but with a different
set of features. I don't know if expanding into
the Schema debates is necessary although it is clear
that we have been conflating these requirements
by discussing expansion of the xml: vocabulary for
things like ids.
Will the SOAP folks be happy with a 'syntax-only' subset?
From: Jeff Lowery [mailto:Jeff.Lowery@creo.com]
The XML I've used:
1. elements + attributes + data content [text nodes]
3. internal entities
4. XML Schema
Obviously, this is no one's subset. Of these, I would say 1 is essential,
2 is necessary. I don't like the namespace prefix mechanism, and I would
probably support the radical solution of designating an official namespace
prefix registry, so that the prefix is synonymous with the URI. You might
be able to get rid of Namespace URI's in that case, or at least mandate that
such URIs, if they exist, refer to a RDDL doc. I don't expect that to be a
popular suggestion, though.
3 could be replaced by a non-DTD declarations. External entities can, also,
though I don't use them myself (at least the DTD versions).
Number 4 is definitely a layer on top, although a purely structural
metalanguage (no types, just form) might be feasible for core. Something
like a simplified DTD, but using element/attribute syntax. Even RELAX NG is
too much, I think.
PI mechanisms can be replaced by a targeted processing vocabulary in XML
syntax. It could be useful if XML components could be designated as "skip
parsing" (like CDATA), unless it's the target application. I'm not sure
about that, though.
I agree with Gavin about Infosets: NIMBY.