Re: XPointer

Roy Fielding wrote:
>The interpretation of fragment identifiers is media-type specific.
>That does not mean that it is a good idea to define the format of
>a fragment differently for every media type.  Allowing the same
>identifier to apply to multiple media types isn't always possible,
>but it is desirable except in those very rare situations
>where the user actually intends to identify a specific XML element
>(as opposed to the content encapsulated by that element).

Which can be regarded as a good support for the fact that the 
proposed framework supports "bare names" as an addressing mode, 
defined to work in the most natural way for XML documents.

I'd argue that the W3C framework explicitly supports the case that 
you are worried about (one where the media type of a resolution is 
unpredictable), while also supporting the important case of 
more-precise addressing when the media types _are_ more predictable.

the W3C has also made no recommendation to the IETF at this point 
about whether the more-complex XPointer pointer scheme should be 
optional, recommended, or required.

So at least part of your objection seems to be to a proposal that has 
not yet been made, and may never be made. I personally don't favor 
making the most-complex pointer schemes mandatory for applications. I 
think that the notion of schemes itself is actually very good, as can 
be judged from the fact that people who are not fans of the XPointer 
scheme are already proposing alternate schemes in its place.

The fallback semantics of the XPointer framework provide a nice way 
to deal with the uncertainty inherent in having a per-media-type 
addressing syntax that must be packaged along with pointers (URI 
references) that may return data conforming to a number of media 
types.

Without such a scheme, it's very hard to create URI fragments because 
any proposal must work across an unbounded number of media types. 
Adopting this as a practical principle, furthermore, means that no 
media type can allow any addressing modes that are unique. You may 
thin XPointers are not an ideal solution for XML documents, but the 
same arguments would prevent indexing multimedia presentations by 
time code, or spatial extent, because those addressing modes would 
not be valid for other media types.



    -- David
-- 

David G. Durand
VP, Software Architecture
12 Bassett St.
Providence RI, 02903-4628 USA

T: +1 401-331-2014 x111
T: +1 401-935-5317 Mobile
E: david.durand@ingenta.com

Received on Monday, 11 November 2002 14:58:17 UTC