[
Lists Home |
Date Index |
Thread Index
]
"Cavnar-Johnson, John" wrote:
> I disagree here. I don't see a fracture between users of validating
> parsers and non-validating parsers today. What's the basis for your
> assertion that this will have serious consequences? Nobody's saying that
> the documentation world can't keep there current tools and methodologies.
> I think the theoretical users of the new parsers have little or no
> interaction with that world anyway. I hate to see this false dichotomy
> between doc heads and data heads perpetuated. I'm interested in systems
> that integrate business documents (purchase orders, contracts, etc.) with
> relational systems. I consider myself very "document-oriented", but DTDs,
> entities and such don't buy me much of anything and the non-optional
> nature of the internal subset is extremely frustrating.
In the fifth anniversary celebratory essays there flickers still the dream
of data fusion, under which there is in fact no dichotomy of doc heads and
data heads. What has changed is the orthodox opinion of what that fusion is
to be built upon. In the 'remembering the original XML vision' thread Dare
Obasanjo succinctly made the case for fusion built upon infosets which, like
you apparently, he believes is the new orthodox opinion. I do not question
that may now be the opinion of the majority. However, I build (and have for
quite some time) 'systems that integrate business documents (purchase
orders, contracts, etc.) with relational systems' and with transactional
processors. I can demonstrate with copious evidence that integration based
upon infosets does not work for the business documents which are in fact
found, as documents, in daily use while conversely integration based upon
syntax and a bare minimum of syntactic rules, like well-formedness under the
XML 1.0 Rec, does work spectacularly well for such integration.
It is unfortunately clear that many are not interested in the empirical
evidence of how well integration upon syntax works, preferring from polemic
motivations to dismiss its details with no better argument than that they
'don't buy me much of anything'. Do you understand how thin and petulant
that argument must sound when what is at stake for me is losing the working
integration which I have already achieved? If document creators follow your
prescriptions their documents become increasingly opaque to me and to the
methods which now work very well in integrating those documents into
transactional processing. When you reach an entirely infoset-based
integration you will have succeeded in limiting its benefits solely to the
cartel who accept your precise premises in toto before ever gaining useful
access to your 'documents'. You will have also, not incidentally, created a
model of 'business document' suitable for your processing which barely
resembles documents which have natively arisen within businesses and remain
the basis of conveying most information there. That is, instead of quick,
easy and working integration of what is now in use on the document side with
the processing methodology which pervades the data side, you will find
yourself offering business only the enormous cost of building your preferred
infrastructure before they ever enjoy the integration you offer, which will
be an integration only with others willing to submit to the same onerous
conditions.
Respectfully,
Walter Perry
|
