Re: which layer for URI processing?

On Wed, 24 May 2000, Tim Berners-Lee wrote:

> This is *NOT* a considerable amount of effort, this is calling a
> simple well-defined string function which takes two string parameters
> and returns one string value.

If you could make that argument, you might be on the same page with more
people.  Unless you have the parser absolutize all namespace references
and disallow the user from doing anything but absolute references through
the DOM (a little bit like the forbid option), it is a big complication
to the object model to have to deal with relative namespace referencing
capabilities within a live model.

And for what?  xml:base lacks the precision to be useful in most cases,
because it fails to distinguish between content location and namespace
identity as separate dimensions.  If anything, the namespace URI would
need to be relative to the DOCTYPE or Schema URI, and the URIs of the
namespaces of surrounding elements -- not relative to the document URI
or the URIs of surrounding content, which seems rather nonsensical for
most use cases I have dealt with.

Why not step back and look at your requirements, rather than the solution
you are advocating, and structure something that meets more people's 
requirements and allows better flexibility for the relative case, without
making people pay for it that don't need it?

Ray Whitmer
ray@xmission.com

Received on Wednesday, 24 May 2000 14:52:36 UTC