[
Lists Home |
Date Index |
Thread Index
]
> > There's not much more I can say to that. There may be occasional
> > exceptions. Data model by decree works occasionally - XPath 1.0, I
> > think is the strongest case in that direction, though it succeeds in
> > large part by doing a lot with very little.
>
> IMHO the XPath data model (also the RNG data model and the Infoset) are
> more like "abstract syntax models". They abstract away a few syntactic
> details, but don't introduce any semantics of their own -- they are
> still couched in terms of elements, attributes, etc.
Seems to me there's a gap in the bits-over-wire line of argument a bit
further down. If you exchange concrete syntax then what you have exchanged
is concrete syntax. Ok, you may have an abstract syntax at either end in
which to interpret that, elements, attributes or whatever. Arguably a
grammar is a data model in itself, but even so, what use is a grammar if
there's no further model being shared? The sharing might not be complete,
the local languages may only partially intersect, but without those semantic
intersections, what exactly is being shared? You say potato, I say potato.
potato potato...
In the context of the web, I suspect its success is due more to shared
abstractions like URI -> representation than syntax. Maybe it's not
expressed explicitly anywhere, but that looks very like a relational data
model. Is a document merely syntax? These things may be simple, but without
them you might as well just probe your network card with an oscilloscope.
Ok, so in a roundabout way that's what my browser is doing a lot of the
time. Hyperlinks could just be considered routing codes. The web can
certainly be reduced in a Newtonian kind of way to these bits on the wire.
But I don't think we'd have the web as it exists today without layers of
abstraction on top, and I think it would be fair to call a lot of those
layers data models. By talking of http and html and so on as purely
syntax-based systems, seems like not seeing the wood for the trees.
Or am I missing something fudamental here - prithee tell, what exactly is
syntax-based interoperability? What messages can be communicated by syntax
alone?
Cheers,
Danny.
PS. 01010011010011110101001100100001
|
