Re: Personal reply to Edd Dumbill's XML Hack Article wrt W3C XML Schema

["W. E. Perry" <wperry@fiduciary.com>]

Richard Tobin wrote:

> Are the simple types sufficiently uncontroversial to form a basis for a
> common "typed infoset"?

I would argue that their very specification violates the unique nature of
XML. In my (apparently unorthodox) view the unique property of XML is
that it places the instance before the type. Said differently, XML is
intended to be eXtensible through Markup in the instance, not through the
extension of a schema, DTD or other data model. The uniquely XML notion
of
well-formedness as the common baseline of conformance should guarantee
this primacy of the instance. By this argument, simple types (and indeed
any data schematics) are as derived from the instance document for the
execution of particular processing on a given occasion. The 'validity' of
the data is a determination that it satisfies the requirements of
processing on that unique occasion. From that perspective, the 'typed
infoset' elaborated from the instance data is not 'common', in the sense
that that it can predictably be elaborated on other occasions from
similar data instances. There is however an 'instance' typed infoset
which might be elaborated from the specific processing of any XML
document on a given occasion and which might usefully be the basis for
further processing. A sequence of such instance infosets could be
pipelined between processes which might be coupled in unique sequences in
order to accomplish data (or document) processing specifically suited to
the circumstances in which it is executed. Specifying data typing which
depends in this way upon the instantiation of instance XML on a given
occasion to meet the expectations of a given process is admittedly a
radical departure from the philosophical premises on which the Schema WG
(and other WG's whose work is alleged to be too dependent on the Schema
WD) operate. It is nevertheless a method for achieving schema
specification--and, I would argue, the only method--which does not
violate the salient point of well-formedness as the unique baseline of
XML conformance.

Respectfully,

Walter Perry