Re: fragment identifiers and media types (was RE:XPointerconsidered incomprehensible)

Dan Connolly wrote:
> On Wed, 2006-09-06 at 15:47 -0400, Booth, David (HP Software - Boston)
> wrote:
>>> From: Dan Connolly [] 

> On balance, I still find # URIs a lot more straightforward than
> redirects.

I was wondering how to make these kinds of #-/ disputes more measurable.

(I didn't succeed...).

How many media type specifications actually define a semantics for what
# "views" mean and/or refer to?

How many of them note that, for any given piece of content that's to be
published in the Web, it shouldn't be deployed in a way that #something
is given different meanings by any of its published content types. If
the PNG spec and the GIF spec and the JPEG specs disagree on what #blah
means, I should avoid making an image available content-negotiated in
those formats. It isnt just a matter of my own self-discipline in not
using "myimage#blah" URIs, ... since the very fact that myimage is
available in GIF, PNG, JPEG or whatever invites folk to read those file
format specs, and ... figure out what #blah means for content of that type.

If most media types don't bother saying anything precise about #blah, I
guess it's not a big problem for content negotiation. But right now
there seems a big question mark about what kinds of document it is
reasonable to put into a content-negotiated bundle: can I sensibly do
this with PDFs and Open Office docs, for example? How would I find out?

For the restricted case of RDF docs, ... I can see that # might win out.
But I'm still left with a big sense of 'uhoh' re use of # identifiers,
because of the many other media types who don't say anything very
precise about it.


Received on Wednesday, 6 September 2006 21:57:35 UTC