Re: Separate treatments for System ID and others URI-References

[Daniel Veillard <veillard@redhat.com>]

On Wed, Aug 22, 2001 at 04:26:56PM +0100, Tony.Coates@reuters.com wrote:
> 
> On 20/08/2001 21:39:01 Daniel Veillard wrote:
> 
> >I really don't see why two similar URI-Reference (assuming similar base)
> >may lead to different resources when processed by the same software and
> >coming from the same entity. This seems to break the very generic
> >nature of URI-references and I really can't see the point, maybe
> >you can provide a rationale for this.
> 
> When the W3C decided that namespaces should be indicated by URIs (ostensibly not related to physical resources), it decided that URLs could be used indicatively rather than as labels for particular resources.  XML Catalogs similarly understand that the same URI may be used to different contexts to represent things.  A namespace URI can indicate a namespace, but it can also be indicative of a default Schema or a default stylesheet, etc.

  Well, the case you point out is a case where the string of an URI-Reference
is used as a string. However it did not change the fundamental algorithm
when using an URI-Reference to retrieve a resource. And XML Catalogs introduce
a divergence in this fundamental algorithm saying it separates an URI-Reference
used as a system Identifier and an URI-Reference used an an URI-Reference :-)

  Practically XML Catalogs changes the XML Recommendation by saying that
the System Identifier is not a normal URI-Reference ... Some people have
complained *very* loudly for less than that.

  This would at least requires an explanation of the difference in the
specification (where I secretely hope I would get the explanation of what
I'm misunderstanding, because it's likely I am missing something).

   Yours,

Daniel

P.S.: that said the next version of libxml/libxslt will embedd a full
      implementation of XML Catalog from 06 Aug 2001, but I'm seriously
      wondering how and why I should use 2 different algorithm depending
      on where an URI-Reference is located. Basically you need to export
      some of the parser internal context into the resolver (or duplicate
      all the APIs), and I honnestly don't see the point of separing
      URI and System ID, you can consider this an formal implementor feedback.

-- 
Daniel Veillard      | Red Hat Network http://redhat.com/products/network/
veillard@redhat.com  | libxml Gnome XML XSLT toolkit  http://xmlsoft.org/
http://veillard.com/ | Rpmfind RPM search engine http://rpmfind.net/