Lists Home |
Date Index |
> And here, I believe, we arrive at the nub: of the Three Abominable XML-DEV
> Permathreads, we are not really engaged in an iteration of Syntax vs.
> Semantics, but of The Proper Processing Model. Jeni and Patrick may make
> assertions about data models and about syntax, but all of their examples
> immediately to their different forms of processing. Is there any
> that 'process . . . takes you from the syntax to the data model', or as I
> usually phrase it in the permathread, semantics are elaborated from syntax
> an instance execution of process? If we agree on that much, then we can
> accept either Sam's assertion that there is no data model, or Jeni's that
> there are many. Between syntax and model lies the execution of a process,
> likewise between a model and some serialization of it. I will continue to
> insist that syntax, or more exactly a concrete instance expression of
> comes first. Others may prefer that the fluid preverbal Gestalt prompt
> such utterance.
> Then "Keith W. Boone" wrote:
> > I think you miss the point of Jeni and Patrick's work. The Syntax of
> > not allow for overlapping markup, something Jeni, Patrick, myself and
> > have to deal with on a regular basis. LMNL and JITTs both provide
> > to resolve those problems.
> To which Walter responded:
> I may be doubly obtuse, but I do not get what point it is that you think I
> missed. My point in the posting to which you have responded is that, as
> thread has developed, it has proven to be about processing models rather
> about the larger relationship of syntax to semantics. Is that not
> what you have said just above?
IMHO, what was originally discussed is how the syntax enables the processing
model, not describing "The Proper Processing Model" [your caps]. Later
disagreements seemed to be about whether LMNL was a syntax or data model.
In fact, I don't believe there is a single Proper Processing Model, just
some that work better than others for a given task. LMNL and JITTs seem to
facilitate better processing models without necessitating XML Syntax.
I happen to like the point of view that LMNL describes a data model.
Finally, I would disagree that "process ... takes you from the syntax to the
data model". Data models can be manipulated directly, irrespective of
process. Syntax can bring you to the data model, but unless you've started
with your data model first, you won't come up with a syntax that makes sense
[or that can necessarily parsed unambiguously via some process].
> > They've both extended the notion of markup in such a way as to provide
> > missing capability in XML, the ability to use overlapping markup. Sure,
> > could develop an XML syntax that models their data in such a way as to
> > for an XML serialization of it, but that ignores the human requirements
> > being able to easily interpret and edit that markup [remember, none of
> > important without humans... ;-)]
> As briefly described above, I have in fact developed such a
> which does respect the well-formedness constraints of XML 1.0. I have
> specifically done what I have done in the attempt to insure that both the
> syntactic input and the semantics elaborated by the execution of process
> immediately comprehensible to humans.
I'd love to see some examples. In my experience, all attempts to model
overlapping markup using XML syntax have failed the vi-test [it can be
easily created and (somewhat less easily) maintained using a text editor].
LMNL and JITTs provide syntax examples that pass that test.