[
Lists Home |
Date Index |
Thread Index
]
james anderson wrote:
> Bill de hÓra wrote:
> > Which 'the abstract data model' would that be?
>
> the one in terms of which operations in your processing-environment-of-choice are
> expressed.
Petitio principii, I fear. The larger point is that after 5+ years of XML processing a
significant number of practitioners have discovered (or reaffirmed what they already
knew) that XML applications are effectively monolithic from the syntactic instance to
the idiosyncratic output. That is the inescapable consequence of a syntactic
definition of XML--a hard fact which has been lost on, or ignored by, ancillary
specifications that insist on introducing specious abstractions between the
syntactically-conformant instance and some processing of it appropriate to a desired
specific outcome. To induce from the nature of particular processing some abstract
data model which appropriately describes an XML input instance as handled by that
process may be a useful exercise in designing or refining the implementation of that
process, but it is at best otiose to the XML instance itself, or indeed to other
processing which might usefully manipulate that instance for other purposes. Grasping
for such abstractions is the fundamental misconception (some days I am tempted to
think 'obstinance') which regularly diverts conversations in the XML community from
XML to not-XML. (Parenthetically, I am *very* sorry that Tim Bray might have absented
himself from this present thread, addressing as it does core issues raised by our
syntactically-defined standard. Given their fundamental nature, raising those
questions is never out of order, nor is stare decesis in the name of the Namespaces
Rec an adequate disposal of them.)
Respectfully,
Walter Perry
|
