[
Lists Home |
Date Index |
Thread Index
]
Don Box wrote:
> In case it wasn't obvious, that sentence should have read "What good are
> datatypes if one CANNOT work in terms of the value space..."
Why, yes, that is precisely the question. Or, phrasing it from the opposite
perspective, we might (should!) ask why XML would accept datatyping based on
anything other than the lexical. The nature of XML is lexical (it's text).
The advantages of XML are lexical (system independence, data permanence, and
interoperability are predicated on acceptance of a syntax, not a data model
nor a processing strategy beyond parsing for well-formedness). Do what you
please locally in your particular rendition of 'value space', because of
course you have to instantiate your data to a local understanding of it if
you are to do anything useful for you with it. Just don't impose that
understanding on your correspondent or counterparty with whom your a priori
agreement extends only so far as XML syntax. The compelling argument for
XML, like the compelling argument for an internetwork topology over an
homogenous enterprise network, is that you can address, and perhaps perform
useful data interchange or transaction execution with, imperfectly-known
counterparties with whom your only shared assumptions are lexical.
Defending, developing, and promoting XML has got to be about preserving that
core premise. Preserving that premise has got to require, at times,
rejecting extensions of XML's common denominator of agreements into
features--however apparently useful--which constrain the possibilities
inherent in the lexical basis. As far as I can see, it is always an easy
test whether some appealing feature is of its own nature in the lexical or
the value space. If the feature is based primarily upon a model rather than
a text or if it doesn't have an obvious 'XML serialization', reject it
because it is not XML. And if you need that feature and the model it serves
more than you need to preserve the lexical basis and possibilities of XML,
well then you have made it clear that the feature belongs somewhere other
than in the domain of a markup specification predicted upon lexical
agreements.
Respectfully,
Walter Perry
|
