Re: meta-specs (was RE: A few things I noticed about w3c's xml-schema)

[Tim Bray <tbray@textuality.com>]

At 06:10 PM 30/05/01 +0100, Sean B. Palmer wrote:
>> Then RDDL is a catalog of relationships among components
>> of some system.
>
>But it only works on the schema level - not on what's inside the
>schemata.  For example, it could point to two schemata, one in TREX,
>one in XSD. One could say that a certain element is allowed in place
>x, and the other could say that it isn't. Which is to be believed?

I'm sure this will happen all the time.  An even more extreme
example is where a RDDL points to two different XSD schemas that 
offer different validation policies.  RDDL offers a weakish
reed to lean on for disambiguation in its xlink:arcrole= attribute, 
but at the end of the day there's going to be no substitute for 
having some human-readable text to explain what these things are 
there for.  That's why RDDL is primarily HTML.

>Then you might have the more subjective layer, which is to say that
>the purpose of this element is y, so go and work out whether or not
>you are allowed to use this in place x. That's the kind of layering
>I'd like to see provided in RDDL somehow, but I'm not quite sure how
>to do it.

This sounds really hard.  You'd need sort of a Universal Processing
Semantics Description language.  Of course, such things exist
(e.g. Java, Perl) but you probably want a declarative one.  -T