Lists Home |
Date Index |
From: "Jeni Tennison" <email@example.com>
> A little while ago, Rick Jelliffe issued the challenge:
> > Personally, I think it is high time for people who don't like the
> > W3C datatypes to band together to make a concrete alternative.
> I thought I'd spend a little time on this, so I've put together a
> "datatype library language". The idea is that applications such as
> RELAX NG validators or XPath 2.0 processors could read in a document
> in this language and use it to construct a datatype library or set of
> in-scope type definitions.
> Note that this isn't a proposed datatype library -- it's not a
> replacement for the XML Schema datatypes -- rather, it's a language
> that enables users to create declarative definitions of the datatype
> libraries that they need.
> A description of the language itself and a proof-of-concept
> implementation are available from:
> I think it's at a stage where it's complete enough to be worth
> discussing but not so complete that it's not worth discussing. In
> other words, I invite comments!
I'm still digesting this. It's very interesting (!), chock full of good
ideas and deserves more careful study.
The only thing that jumped out as a potential issue, which you also
highlighted, is the incorporation of XSLT into the language. I'm afraid this
is the camel's nose in the tent. Though your use of XSLT in the examples is
carefully restricted, leaving the door ajar will invite the full language in
and effectively prevent any implementation other than by XSLT. I am
concerned that a process that might be performed as a side effect of
validation could have overheads that swamp those of validation, itself.
My suggestion is to carefully evaluate where you found XSLT useful (as in
providing a template capability for error messages) and design your own
language around these requirements. Any implementation via XSLT is going to
first translate the document into an XSLT stylesheet, anyway, and it will
have no trouble translating those elements (back) into XSLT syntax.
Unlike Dare, I see no problem with a default namespace. This works well in
RELAX NG and is especially handy when including fragments developed or
"repurposed" to be used with multiple namespaces.
Hope to have more constructive suggestions after another pass.