[
Lists Home |
Date Index |
Thread Index
]
Michael Kay wrote:
> Secondly, the push model actually binds the stylesheet less tightly to
> the structure of the source document than the pull model.
I regularly make this point, only to find that it is disregarded or
doubted, so I am very pleased to have your authority for it. In the
processing topology which I describe, this tight binding of the stylesheet
to the source document is highly desirable. Either that document itself
will have been sent to a processing node, or that processing node must
pull a source tree from another internetwork node by specifying
element-by-element precisely what is required. In the latter case, the
node asked to supply data needs a precise target form into which to
instantiate what it exports; tight binding of source tree to stylesheet on
the requesting node assures that the request is sufficiently precise. In
the alternate case, where a document is sent to a processing node, that
document is itself the most precise representation of the source form
which must be instantiated into the specific (target) structure required
for the execution of local processing. The generic facility which I
advocate for performing that instantiation works on the principle of 'from
what you actually have to precisely what you require'--a very tight
binding on both sides of the transformation. A practical requirement,
first of parameterizing and then of executing that transformation, is that
particular instance data of both the source form and the target form be
available for sanity checks of the instantiation. In other words, the
transformation which performs the instantiation must be tuned on both its
input and output sides by the specific instance data at hand. If that
transformation is performed at the node receiving a document, that
document itself answers the requirement for tightly bound source, just as
a transformation performed by a node supplying data in response to a
'pull' request can rely on the form of that request to be the precise
statement of the form of data required.
Respectfully,
Walter Perry
|
