On Thu, 24 Jan 2002, Mark Baker wrote:
Personally, I believe that the root namespace is not a definitive guide
to the contents of the document.
I agree. It *can* be used as an indicator of the first processor to
be handed the document (after the one dispatched from the media type),
but it need not as we've seen.
FWIW, my new media type will require that content expects that it will
be processed in this manner. But */xml and */*+xml don't have any
such requirement.
IN retrospect I realize that XML content-types really just define a 'most
important' namespace, in the sense that they reference the first layer of
'xml-aware' software that should handle the resource.
Following on this model, I find that Simon's draft makes more sense (to
me) if I think of it as defining the 'relative importance' of different
namespaces in the data, particularly when 'q' values are used to define
quality parameters.
Looking at it this way make the content-features header a more natural
extension on information provided by the content-type.
So, in a sense, this takes the set of namespaces in a document (root + all
others) and maps them into a different set that weights the different
namespaces in an application-specific manner....
Ian