[
Lists Home |
Date Index |
Thread Index
]
"Simon St.Laurent" wrote:
> Are there more issues like Character Entities waiting to surface? Or
> (apart from that issue) can XML declare victory and call itself complete?
Good question. In my opinion the answer depends on whether we have exhausted
ourselves with arguments over processing models or are, in fact, finally
going to face up to defining in XML specification terms what happens after
the parse. The mention, and the promise, of an XML processor clearly seen as
something beyond the parse appears throughout the XML 1.0 Rec. Except for
WXS and XSLT, the XML technologies which you review are all about aspects of
the parse. I regard them all as effectively marshalling functions: various
aspects of processing which might be invoked to prepare something very like
a stream of SAX events to move on toward the 'real' processor. Some of these
technologies certainly appear to come earlier or later in that marshalling.
The character entities discussion essentially foundered on the unyielding
rock of missing entity declarations being a well-formedness rather than a
validity error. That is, the problem is identified, and causes its
consequences, too early in the marshalling process for something like your
ents processing to be its recovery mechanism. XInclude arguments typically
turn on where in the relative order of marshalling the includes are to be
executed. Even the recent HLink/XLink brouhaha was in large part about the
use of the structural or addressing (presumably earlier marshalling) versus
presentation (presumably later) feature of the two proposals. The mismatches
of XPath syntax for what XQuery wants to achieve come down to the same
questions of what takes precedence in the marshalling, even if the arguments
are usually couched in terms of abstract document/data models. Even WXS is
really about functions which might be invoked in marshalling, even though
they seem clearly to come very late in the process, providing final
decoration to produce a PSVI. And XLST is so closely tied to the specific
form in which the output of the parse is marshalled that, though it might
provide processing beyond the parse, it does not do it at very much of a
distance.
I agree with your implication that we may have gone about as far as we can
in tinkering with the marshalling functions aggregated in the parse. I do
think that the next job--and one which may give us enough perspective to
solve some of the remaining irritants and miscoordinations of the
marshalling components--is to look at the de-marshalling that needs to be
part of the post-parse processor.
Respectfully,
Walter Perry
|
