Re: breaking up?

["Simon St.Laurent" <simonstl@simonstl.com>]

On 05 Aug 2001 10:46:40 -0700, Tim Bray wrote:
> At 10:59 AM 05/08/01 -0400, Simon St.Laurent wrote:
> >On the other side seem to be people who find the results of XML 1.0 interesting but not nearly good enough by themselves.  These folks seem intent on decorating XML with a number of features - Namespaces, W3C XML Schema, XInclude, etc. - which add to XML's capabilities at some cost in its clarity.
> 
> Hmm, do you claim that all the things that have been piled on
> XML 1.0 fall into a single qualitative bucket?  It seems to me that
> Namespaces, XSLT, and XSchema (to pick 3) are horses of very different 
> colours, in terms of scope, philosophy, and general intellectual space.

Not necessarily a single bucket, but a common point of departure.
There's a big gap between explicit-in-markup and
added-in-uncertain-metadata.

XML 1.0 already straddled that gap (thanks to DTDs), but namespaces and
XML Schema widen it considerably.

Sort of a document-says-it-all vs. infoset-says-it-all breakdown,
complicated by other divides (document/data, object/RDBMS/document,
etc.).

> To me, the fragmentation point is between those who think the 
> data structures are the real thing - the Schema/Infoset/PSVI/Query
> world-view - and those who want to maximize interoperability at
> the level of syntax: XML/namespaces/SAX/maybe-XSLT.  And I agree
> with Simon that this fragmentation is not necessarily damaging.

Not necessarily damaging, but worth acknowledging in ways that make
explicit that moving forward into Infoset involves abandoning many of
the priorities that led to XML's creation and initial exitement.

> I have to say that a high proportion of the real-world apps
> of XML I see are concerned with generating tags & attributes
> at one end of an interface and parsing them at another, and
> the people writing them never think for an instant about infosets
> or subelement qualification.  And hand-creation and on-screen
> viewing of XML are quite common, but mostly in design & debug
> mode, which seems appropriate. -Tim

True, though I'm hearing more and more from people who are encountering
the "missing data" issues I've been pointing out for years.  Problems
discussed on this list periodically manifest themselves in reality as
well!