Lists Home |
Date Index |
On Tue, 30 Aug 2005, Bullard, Claude L (Len) wrote:
> This is fun. Can anyone here describe the hypothetical
> most stable turtle stack keeping in mind that losing it in the
> philosophy of meaningfulness is not being a clean
> clear thinking turtle? The first turtle is the human
> who is the ultimate semantic processor - the authority -
> so it means what you say it means and if you can't
> communicate your meaning, you are a fuzzy turtle.
> What is the ideal XML application language architecture?
> Is there such a thing? Are we stuck in a world where
> all we can rely on is the syntax (the Perry Position)?
> Can one layer in the semantics cleanly and clearly
> in such a fashion that the techniques are sharable
> and the semantics are fully learnable or discoverable
> with emphasis on "learnable" because discovery
> infers one is uncertain whereas learning means one is
> just scheduled?
Doesn't the following stack fit the bill?:
Top: 7. Humans
5. High-level programming languages
4. Declarative invariants
3. Types (i.e., structural constraints)
2. Abstract syntax [XML Infoset, or other XML data models]
Bottom: 1. Concrete syntax [XML]
This is just an XML-specific instance of the stack common to software in
general. The stack became more refined over the decades: originally all
semantics was in assembly-coded applications (#6), but gradually the
layers #2-#5 have separated.
Unfortunately, XML practice doesn't align along this stack cleanly. (I'd
say, this is too bad for practice ;-) E.g., W3C Schema spreads across
several layers, tightly glueing them together: it hijacks #2 by putting
PSVI instead of the Infoset, it's bulk (rightly so) is #3, but it also has
things belonging to #4 (e.g. key/keyref constraints) and even #5 or #6
(e.g. the mechanism for filling-in default values and the mechanism for
translating XML into PSVI).
Or you were asking about something totally different? If you have stuff
like RDF in mind, it is an application that uses XML as just a substrate
on which it grows. (I.e., RDF has a similar stack, and if we care only
about XML-based RDF, the interesting part of this stack is inside #5 and
#6 above.) But RDF, or similar "semantic web" developments, can't be the
only way to give XML semantics! (Sorry if the RDF guess was wrong)