Re[2]: Summary: Section 2: What does a URI identify?

On Tuesday, 19 March, 2002, 09:38:25, Brian wrote:

BM> At 14:20 18/03/2002 -0600, Paul Grosso wrote:
BM> [...]

>>No!  People keep getting this wrong, and unfortunately major browsers
>>do too which is really making a mess of the use of base URLs to the
>>point where one cannot use <base> and <a href="#xxx"> sorts of references
>>together, because this misinterpretation of #xxx refs with respect to
>>base URL destroys the ability to maintain intradocument xrefs.

BM> That looks like an eminently reasonable requirement for HTML.

That seems like marginalization to me - the HTML WG is clearly working
to make XHTML be generic XML, and the same methods should apply to
XHTML, SVG, MathML, SMIL, fooML and barML when there is xml:base and a
link with a URI that is just a fragment identifier.

The same interpretation should apply in all cases that use xml:base
and that interpretation should be in the xml:base spec, not in any
spec that is a client of xml:base.

BM> Let me explain what RDF is trying to accomplish.

BM> For good or ill, RDFCore has inherited a situation where:

BM>    <rdf:Resource rdf:ID="foo">

BM> names a resource whose URI is <base-uri>#foo, i.e. it is relative to the 
BM> document containing the element.

(Request for clarification - the rdf:ID attribute is of type ID in the
DTD or Schema?)

BM> These relative references cause problems.  One of the most frequent newbie 
BM> errors with DAML+OIL I have seen is "I have tried processing 
BM>[*], and it didn't work right."  The reason 
BM> is that the user should be using 
BM> contains 
BM> exactly the same text, but because it contains these pesky relative 
BM> references using rdf:ID, the  URI's generated by the parser are wrong.

BM> I first came across this problem when I copied the RDF schema to my hard 
BM> drive so I could work on some code on a plane.  I was a little surprised to 
BM> find I had uri's of the form:

BM>    file:c:\temp\rdfschema.rdf#Class

Its clear that any situation where something works online but a
local(y accessed) copy suddenly breaks is a problem. The XML MIME type
RFC, by introducing a precedence of headers over the XML encoding
declaration, has a similar problem.

BM> What is needed is a way to specify, in the XML document, a base for 
BM> converting relative URI references to absolute form.  Then it would not 
BM> matter, where the file was accessed from.

BM> The other obvious use case is where the RDF in question has no obvious URI, 
BM> e.g. when it is generated as part of protocol, perhaps for e.g. in cc/pp.

BM> Whilst the RDF community could invent its own mechanism for this, further 
BM> divergence of RDF from XML does not seem like a good thing.  We have 
BM> decided therefore, to test whether we can use xml:base for this purpose.

That does seem to be the purpose of xml:base. However the MIME
registration for RDF gets to define fragment identifiers in whatever
way is appropriate.

BM> It seems that HTML and RDF have different goals here.  I wonder if there is 
BM> a way to reconcile them.

Lets not restrict this to "HTML vs RDF" but try to see a wider picture.

BM> ps: It would to keep clearly separate in our minds, the issue of what does 
BM> a URI (reference) identify and the use of xml base.

I don't see how it is possible to separate them. The meaning of the
latter is intimately dependent on the former.


Received on Tuesday, 19 March 2002 10:39:20 UTC