Re: Excess URI schemes considered harmful

It's not a direct mapping; media types are used for dispatch and
identifying the document, whilst namespaces are used for identifying
(but not dispatching) XML elements.

As a result, it's many-to-one; in time, there could be several
versions of SVG (and hence namespaces) mapping to image/svg+xml.

This is a somewhat desireable effect of using media types; not
disasterous to lose, but worth bearing in mind.

Similarly, the mapping is arbitrary; for example, an RSS 1.0
document's root namespace is that of RDF, yet there is still a need
(satisfied by media type) to identify that document as RSS.

> RFC 3023 suggests content types for several W3C defined languages
> including SVG, RDF, MathML, and XSLT all of which are identified by
> namespace URIs.

I think the question here is whether a namespace URI can be used to
identify a document, for purposes of dispatch, as well as identifying
(some of) the XML elements within. My impression of the namespaces
discussion is that using them for dispatch in XML is overloading
them; it isn't designed into them, but it can be done with careful
thought. Using them as document identifiers would seem to be more of
the same.

This is interesting in light of the XML Processing Model workshop; I
don't think namespaces and media types as dispatch mechanisms weren't
talked about very much.



On Tue, Oct 30, 2001 at 11:18:13PM -0800, Henrik Frystyk Nielsen wrote:
> 
> >> I agree, but this seems very different from the mapping proposed by
> >> RFC 3023 [1] which defines a mapping from a (namespace) URI into a 
> >> content type based on an IANA registration and a new "+xml" syntax 
> >> convention for XML based content type names.
> >
> >Where does 3023 propose such a mapping?
> 
> Through the central registration process. This is for example how the
> "http://www.w3.org/2000/svg" namespace identifier identifying the SVG
> language is proposed to be mapped into "image/svg+xml". This is of
> course not an automatic mapping but a mapping nonetheless.
> 
> Henrik
> 

-- 
Mark Nottingham, Research Scientist
Akamai Technologies (San Mateo, CA USA)

Received on Wednesday, 31 October 2001 12:34:25 UTC