Re: URIs quack like a duck

"Arnold, Curt" wrote:

> Basically since a schema can define a set of constraints for a group of namespaces, it seems much cleaner if you identify the schema resources that govern the document in a particular priority order.
> I'd definitely lean towards something like:
>
>     <foo
>         xmlns:x="uuid:12345"
>         xmlns:y="uuid:67890"
>          xmi:schemaLocation="http://www.someone.org/data/schemaQ">
>          <bar x:a="4"></bar>
>      </foo>

> If I did this,
>
>     <foo
>         xmlns:x="uuid:12345"
>         xmlns:y="uuid:67890"
>          xmlschema:x="http://www.someone.org/data/schemaQ"
>          xmlschema:y="http://www.someone.org/data/schemaT"
>          <bar x:a="4"></bar>
>      </foo>
>
> and schemaQ imported a different schema (schemaV) for namespace uuid:67890 which one do I do to resolve:
>
> <x:foo>
>         <y:bar/>
> </x:foo>
>
> If y:bar is validated using schemaT, then I violate the expectation of schemaQ which only expected y:bar to have the forms allowed in schemaV.

In other words, you're saying that attaching schemas to namespaces via the xmlschema:foo attribute isn't going to work, right?

What I wonder about is this.  The namespace spec says:

"It is not a goal that it [the namespace name] be directly usable for retrieval of a  schema (if any exists)."

>From this I infer that using the namespace name for direct retrieval of a schema (which the xmlschema attribute does and xsi:schemaLocation doesn't) is perfectly plausible; the namespace spec writers
just don't have it as a goal.   If it weren't plausible, the statement would be pointless.

What did they (and some of them are here, of course) have in mind?

Paul Abrahams