RE: [xml-dev] DTDs, W3C Schemas, RELAX NG, Schematron?

["Bullard, Claude L (Len)" <clbullar@ingr.com>]

Fundamentally I agree.  But this comes down to 
what do we mean by "system" and "system scope". 
If we want seamless interoperability of a given 
system with another given system, value type 
standards are very sucessful.  But the ubiquity 
of such?  I seldom encounter such if ever.

In our local systems, we have to support three 
database backends.  For these, we map the value 
space datatypes in code and in the data dictionary 
(where the DD is really a document).  ODBC layers 
take care of some of this and careful decoupling 
in the client does some of this.  We choose what 
we will support based on issues other than standards 
or authority.  Typically, price and performance are 
the first razors.  Oracle is a wonderful performer 
but SQL Server is an adequate performer at a much 
better price break.  So if a value space standard 
conflicted with that, we would ignore it. We would 
have to.  It is the hard reality of competitive 
procurement that having the best system at the 
highest bid is *usually* a loser.

So it comes down to the scope of the interoperating 
system vs the freedom of choice in implementation of 
any local system that must interoperate with all of 
the other "local systems".   This is the dilemma that 
finally forced the W3C to buy into SGML On The Web. 
It loosened the definition of system and strengthened 
the power of "locale".

It is why I say "there ain't no web; just parts 
and assemblies".  Usually when someone disputes that 
assertion, the arguments come down to locus of 
authority to define scope for a system, and typically, 
they mean "all of the sharable information space as 
declared in URIs and governed by the W3C".  

It ain't happening.  That definition is screwed for 
all practical purposes by the requirements on the local 
implementor.   They might want to ramp back to 
something more workable and less ambitious.

Pluggable types with a non-normative primitive set 
seems to be what is best and moreorless how we 
do business in other application domains.  As John 
said, the TAO of RNG.  Regardless of authority, that 
gives it an ecosystem survival edge.

len


From: W. E. Perry [mailto:wperry@fiduciary.com]

IMH (and oft-stated) opinion, we lose the very promise of XML. The number of
correspondents and counterparties with whom we might exchange documents and
perhaps execute transactions is vastly greater if the only required
preconditions are lexical--i.e., the syntax of well formed XML. Granted,
because we lack prior agreements in 'value space' on which to understand
those interchanges and predicate those transactions, it will be necessary
(and often long and painful) for us to build up with lexical tools the
minimal one-to-one understandings we need to give each of those interactions
its necessary meaning. The point is that it can be done, however distasteful
the process for doing it anew with each new correspondent might seem to
those who would rather short-circuit the effort by limiting their
interactions to those who will accept a priori their definitions of 'value
space'.

I have spent more than twenty years working with applications built on RDBMs
and relational wannabes. They are great and useful tools, but only in an
homogenous enterprise network where the processing nodes have intimate
(white box) knowledge of each others' processing and base their
interoperability on working against identical data structures. Those are
precisely not the conditions of the internetwork topology. If we want to
extend the possibility of interoperability to every potential internetwork
participant, we cannot begin by first constraining to universe of those to
whom we will talk to only those who are willing to accept a priori our own
peculiar and limited renditions of value space.