[
Lists Home |
Date Index |
Thread Index
]
Tim Bray wrote:
>
>...
>
> The fallacy that there can be a single right schema or stylesheet
> or whatever is probably the central motivation that drove me
> into the work of helping with RDDL.
Agreed. But given:
<foo xmlns="http://www.prescod.net/foo";>
<h1 xmlns="http://www.w3.org/1999/xhtml";>Blah
<bar xmlns="http://www.prescod.net/foo";>
<h2 xmlns="http://www.w3.org/1999/xhtml";>Blah</h2>
</bar>
</h1>
</foo>
The message above arrives in your XML inbox. Your XML-aware editor wants
to construct an XML schema (or stylesheet, or ...) to guide you through
changes. What precisely would it look for in the RDDLs for XHTML and
"foo"? How will it know not to try using the "standard" XHTML and XSLT
schemas? It doesn't even have to be the "right" schema (or stylesheet,
or ...). It just one that works *at all*. My vocabulary happens to
change the grammar of HTML such that h2 elements are allowed within bar
elements which are allowed within h1 elements. This is legal according
to the namespaces spec because namespaces are just names. And XSLT is a
perfect example of a namespace that does stuff like this.
Now if we treat this as a document, rather than as a collection of
namespaces, then it has some document type (or MIME type, if you
prefer). Associating a schema (or set of schemas) with a MIME type or
document type (or individual document!) is pretty simple.
> I don't think you've established that goal #3 is not achievable
> in the general case. In the face of lousy hodge-podge designs
> that pay little attention to composability, obviously pointers
> to the relevant bits of schema-ware and so on aren't going to
> help. But we can hope for better than that. -Tim
Is XSLT a lousy hodge-podge design? In the early days of namespaces I
thought it was one of the few compelling examples that showed how two
separately defined namespaces could work together. But trying to build a
schema for the "XSLT+XHTML" document type from schemas for XSLT and
XHTML would be crazy. And validation is the simplest possible behaviour.
It just returns a boolean. The problems get even worse when you talk
about rendering, for example.
Once again, finding a renderer for a mime type or document type (in the
loose sense) is relatively easy...and RDDL seems as if it would be more
useful as the referent for a mime type. Then the RDDL for the mime type
could assemble namespace-specifc stylesheet and schema fragments that
are known to work together in the "right way".
Paul Prescod
|
