W3C home > Mailing lists > Public > xml-uri@w3.org > June 2000

In praise of fixed base was: Layering XPath/XSLT namespaces is unacceptable

From: David Carlisle <david@dcarlisle.demon.co.uk>
Date: Wed, 14 Jun 2000 23:21:16 +0100 (BST)
Message-Id: <m132LXA-000OdCC@dcarlisle.demon.co.uk>
To: timbl@w3.org
CC: xml-uri@w3.org

> RDF uses a 1:1 mapping between RDF properties and XML element types.
> You can't have different implementation of identity for these. Paraphrasing
> your point,

The mapping for RDF isn't 1:1 in anycase

{http://www.example.com/x}xx and {http://www.example.com/xx}x 
ie the element names xx and x in the respective namespaces, end up
with the same URI for RDF.

elements not in a namespace are out of scope for RDF (I think? unless
I missed something)

So if RDF can't handle the (very small minority) of elements with
unfortunate namespace names like "bat" then this can be seen as an
acceptable quirk rather than the end of the world.

I do however agree with the original subject line, the namespace rec
specifies what the namespace name of an element is. The APIs for
determining that name (xpath and dom so far) had better give the same

However while I think that the "literal" interpretaion is by far the
most consistent solution (and it is what the namespace spec says, and
has been the consensus on any list that's discussed this in the past)
I do accept that having RDF being able to refer to these elements
is a reasonable wish and may even be sufficient cause to contemplate
changing the spec. The "fixed base" option does provide that I

"fixed base" does 

ensure that every namespace name is an absolute URI (+ frag id)

makes xmlns="foo" and xmlns"./foo" equivalent declarations (in
contravention of the current spec, but I can't believe anyone
really requires these to be different)

Gives RDF a mechanism to refer to every element name in a namespace
(except for the known problems referred to above)

Keeps all documents that  currently conform to the namespace
conforming, unless they have really used ./foo and foo as two distinct
namespace names in the same attribute list. I doubt whether any real
examples of this could be found (although it would have to be
advertised as an incompatibility)

Keeps document/stylesheet pairs that use the literal interpretation
working, as the namespace names in the stylesheet and document will
be made absolute with respect to the same base, so if they were equal
with the literal interpretation then they will be equal after being
made absolute.

Has some chance of closing this list before Christmas.

Received on Wednesday, 14 June 2000 18:16:16 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 6 January 2015 20:32:44 UTC