[
Lists Home |
Date Index |
Thread Index
]
David Frenkel wrote:
> I would also add that just because an XML file is human readable does not
> mean you understand the context of the tags. A tag of <name> could be
> ambiguous with plenty of possible meanings.
I believe that this, while certainly a widely held opinion, precisely and
profoundly misses the point. If you can successfully process an XML instance
it is because you have a context (which pervades your particular processing)
within which that XML instance is understandable with an accuracy sufficient
for the execution of software. That context of yours might be roughly
equivalent to what was expected by the creator of that XML instance, in
which case you might be said to carry out the creator's intent with your
processing. The is the usual and indeed the mandated case in transactional
processing based upon two-phase commit, behind the firewall where uniformity
of context is assured by specific data structures shared between processes.
That approach however affords only a tiny fraction of the possibilities of
XML-as-marked-up-text. With markup I don't need to understand the document
creator's 'context of the tags'; I need only to have a context which can use
for its own purposes the tags and other markup as given in the XML instance,
without regard for how anyone else might use them or what the document
creator's 'intent' might be.
However compelling are the virtues of XML already enumerated in this thread,
for what we do--processing data with software--this quality of
XML-as-marked-up-text is paramount.. This is how the nature of the data
which is processed is separated from the nature of the software which
processes it. Processing software is built on the assumption of, and
therefore demands, very specific data structures upon which to operate.
However if such a data structure, or something structurally identical to it,
is the basis of interoperability between processes then such processes will
never interoperate with other processes based upon radically different
contexts and premises. The promise of the worldwide internetwork is just
such interoperability and the benefit which it offers by harnessing the
specific expertise of processes based upon widely divergent assumptions. The
ambiguity--or more precisely the semantic mutability--of labels as opposed
to the fixity and brittleness of data types is what makes this possible. Or,
if you will, XML as markup rather than as data representation,
Respectfully,
Walter Perry
|
