RE: Namespace or document gloss?

[Paul Grosso <pgrosso@arbortext.com>]

At 22:39 2001 01 02 +0000, Miles Sabin wrote:
>Edd Dumbill wrote,
>> I was thinking along similar lines, too.  As indeed Paul Grosso 
>> observed earlier there are elements of this that fit with the 
>> "XML Packaging" thing.
>>
>> Maybe all that's needed is a way (PI or attribute) for a 
>> document instance to indicate a relevant resource bundle (or 
>> whatever it gets called...) -- or maybe documents can 
>> incorporate inline aspects from the resource vocabulary.
>
>I don't think that's the right way to go. By hardwiring
>references to associated resources into the document instance
>itself you make it impossible to reuse an existing doc with a
>different collection of resources without modifying the doc, or
>using some XML processor specific override mechanism (eg. a SAX
>entity resolver).

There is a fallacy somewhere here (and I don't mean to pick on
Miles, but his comment above reminds me that I've seen this 
elsewhere too).  If we are talking about standardizing something 
in the belief that doing so is a solution to anything, then 
how can we suggest that an entity resolver--as already 
standardized by TR9401 [1], as XML-ized in Cowan's XML Catalog [2], 
as implemented in various places including publicly available code [3], 
and as being re-standardized in the OASIS ER TC [4]--is somehow
inappropriately "XML processor specific."  Granted such an entity
resolver isn't part of XML 1.0 (I wanted it to be and even wrote
some text to go into XML 1.0, but the XML WG debated the issue
and we finally decided not to put this into XML 1.0), but neither
would be any other standard being invented today. 

Rather, I suggest that using such an entity resolution mechanism 
is absolutely a reasonable way to address Miles' important desire 
to reuse an existing document with different collections of resources 
without modifying the document, and I'd suggest we all throw our
support behind the OASIS ER TC work and work to support its wide
deployment.  If we aren't going to support the workable standards
that exist or are almost in existence to do entity resolution, I 
see no point in developing a standard to support resource discovery.

paul

[1] http://www.oasis-open.org/committees/entity/9401.html
[2] http://www.ccil.org/~cowan/XML/XCatalog.html
[3] 
http://www.arbortext.com/Think_Tank/Norm_s_Column/Issue_Three/issue_three.html
[4] http://www.oasis-open.org/committees/entity/