Re: URIs for terms: motivation [was: Requirements Document]

Your message is an excellent condensation of some of the most-important
aspects of the WWW.  

Thanks.


Now on to technical points:

1/ I believe that you actually don't want URIs in the requirement, but URIs
   plus fragment indentifiers.

2/ I believe that you actually don't want QNames to be syntactic shorthand
   for URIs, but instead want them to be syntactic shorthand for URIs plus
   fragment identifiers, and, moreover, that the local part of the QName
   corresponds to the fragment identifier.

3/ I want to have some mechanism in OWL that can be used to identify
   (global?) XML Schema declarations.  XML Schema seems to indicate that
   URIs plus fragment identifiers cannot be used for this purpose, so I want
   something more than URIs plus fragment identifiers.  


Why don't URIs plus fragment identifiers work for XML Schema?  Well, I
guess that it is because there can be several portions of an XML Schema
with the same ``name''.  That is, there can be both a global type
definition and a global element declaration with the same name in an XML
Schema document, and both of these are referred to by the same QName.

Is this a bug?  Maybe.  Is it motivated by compatability-with-DTD reasons?
I don't think so.  Will it go away?  I don't know, but I'm sure that there
would be howls if elements and attributes had to have different names.

What can we do?  I really don't know.  All I know is that I want to be able
to incorporate XML Schema definitions into ontologies, maybe not now, but
certainly later.

It does occur to me that there are ways around the problem.  For example, I
don't see a reason that forbids the splitting of the XML Schema URI plus
fragment namespace into different sub-spaces.  To refer to a global element
definition, you would use foo:element.bar and foo:attribute.bar for a
global attribute definition.  This would result in different URIs plus
fragments for the different definitions.  In XML you could still use
foo:bar as an element name and foo:bar as an attribute name.  Would this
cause howls of outrage?  I'm not sure.


Peter F. Patel-Schneider

Received on Friday, 8 February 2002 17:03:49 UTC