dns media type registration tree

[Mark Baker]

Ok, I think we've done a good job boiling the disagreement down to its
essence.  Thanks.

It would be good if other folks could chime in with their views, so we
know where we stand.

Also, I should add that I've not been coordinating with Mark Nottingham
during this discussion, and I don't claim to speak for him.  But I think
our positions are very closely aligned.

And, just to quickly answer this question (we can take it offline if
you like) ...

On Tue, Mar 02, 2004 at 10:51:51AM -0500, Keith Moore wrote:
> > It's not multi-layer; the URI and the media type remain the only
> > dispatch points for requests.  They're just dispatching a very generic
> > application.
> 
> and you don't think there is ever a need for further dispatching at the XML
> level? in other words, all XML documents are going to the same application?

Almost; the URI becomes the sole dispatch point.  It's very similar to
tuple space based systems (e.g. Linda), where the identifier for the
space is all that's used for dispatch.  In my experience, it's a really
nice way to build a wide variety of applications.

I'm also not saying that non-RDF uses of XML (and therefore other
*/*+xml media types) won't be needed, only that in my experience,
RDF/XML is a decent 95% solution.  Hence my comment about RDF/XML
"moderating" the need for an explosion of XML media types.

Mark.
-- 
Mark Baker.   Ottawa, Ontario, CANADA.        http://www.markbaker.ca