RE: [xml-dev] Comparable considered necessary

["Paul Brown" <prb@fivesight.com>]

> -----Original Message-----
> [ Simon St.Laurent [simonstl 't simonstl.com] ]
> I've been thinking about the kind of processing I do with XML.
> Nearly all of it involves matching against patterns.
> I think the largest concrete problem I have with URIs is their
> lack of a common mechanism for saying this equals that.
> [ URL-based example for monasticxml.com ]

I have seen someone angry enough at a non-working XSLT stylesheet to bang their keyboard on their cubical walls; pointing out that the correct namespace was

	http://www.w3.org/1999/XSL/Transform

as opposed to

	http://www.w3c.org/1999/XSL/Transform

was small consolation.  (Pointing an HTTP GET at either of those will get you "Someday a schema for XSL Transforms will live here".)

> The lack of clarity - heck, the outright refusal to acknowledge the
> question - about how to get from an identifier to a resource or back
> again - is the nails in the coffin.

Here's a link to the relevant RFC, for reference:

	http://www.ietf.org/rfc/rfc2396.txt?number=2396

What about the situation when a resource identified by a URI (Uniform Resource Identifier) is abstract?  (This isn't forbidden by RFC2396...  It isn't forbidden, but I could legitimately use "plato:chair" as a URI understood to point to the abstract notion of a chair.)

I think that functional equality, i.e., equivalence of interpretation, of URIs is fundamentally an application-level concern.  What seems reasonable to me is to use the URI "scheme" (section 3 in the RFC) to determine an equality relation on URIs.  The RFC didn't do us any favors by not saying (e.g., "a namespace whose name starts with [Xx][Mm][Ll] is reserved") that the schemes http:, ftp:, gopher:, etc. are all reserved and to be interpreted according to various other RFCs.

For instance, my suggestion for the XSL URI would have been: "w3c://XSL/Transform?version=1_0".

Use of URIs of the form "http://foo/bar"; is too widespread to declare that the equivalence relation for the http scheme is the same in a browser as it is in an XML document.  Maybe something (silly) like: "url://http://foo/bar";?

The idea that "http://www.company.com represents Company.com" is not lost on me, but it is definitely inconsistent with the RFC's intent (as I read it).  The scheme is manifestly the one for http URLs, and it points me to Company.com's web page, not to the company.  Maybe something EDI-like, e.g., "d-and-b:1234" would have been better.

	-- Paul