Re: RDF namespace conventions

> Good.  I had interpreted David Megginsons's message
> as prohibiting that, as it was pretty explicit. All the people who
> suggest using the schema-location attribute seem to be arguing
> that way too.

There are two separate issues that need to be kept distinct.

One is whether or not you should ever dereference the namespace name.
This is, as we all know, "not a goal" but clearly it is allowed and
higher level protocols will do that for documents where they know it
makes sense.

The second issue is whether one should ever use the namespace URI
to reference a schema,instead of using schemaLocation (or some other
Here the argument is that it is almost _never_ correct to so closely
identify a namespace and a schema. The relationship between schema
and namespace could be described as "many-many" or "slight" but
saying that a schema validator should use the namespace name to find the
schema appears to imply a very close association between schema
documents and and namespace names.

If there are two identical copies of a schema document then they define
the same schema, you could refer in a document instance to either one of
them using schemaLocation and the document would validate (if it was
valid) But you could not (usually) refer to either one using xmlns, as 
changing the namespace URI from one copy of the document to the other
takes all the elements in the document out of whatever namespace they
were in. xmlns just has really terrible characteristics as a mechanism
for locating schema.


Received on Monday, 22 May 2000 12:39:50 UTC