xml-dev - Bring back the Semantic Web! (Was RE: interoperability)

[Matthew Gertner <matthew@praxisxml.com>]

Jonathan Robie wrote:
> XML solves the problem of syntactic interoperability. It does not solve
the problem of 
> semantic interoperability. It was never designed to. Frankly, I'm not sure
what a good 
> solution to this problem looks like - has there every been a good general
solution in other
> realms of software?

> Now that XML has made syntax so much easier, a lot of people are very
eager to tackle the 
> semantics problem. But let's not blame XML for the fact that semantics are
hard!

Yes, the advantages of a common syntactic framework are immense, and the
hypothetical advantages of some kind of semantic framework would be a few
orders of magnitude more so. Of course, the difficulties are commensurately
greater as well. I also agree that the standard approach to syntax provided
by XML is a prerequisite to any similar effort in semantics.

I'm not quite so sure about the lack of a clear vision of what this common
semantic framework could look like. Isn't this what RDF is designed for?
(This question can be asked with a straight face in just about any context,
right? :-) Anyway, I don't like RDF because I see too much overlap (and too
little interoperability) with XML Schemas. Look at Extensibility's Schema
Adjunct Framework (www.extensibility.com/saf). Even if you don't like the
specific approach (and I like most aspects of it), the general idea is
extremely powerful. The ability to bind arbitrary metadata to schema
components (i.e. elements and attributes) means that we can formulate the
semantics of a given component in an elegant way at schema level, specifying
for example:

1) How a value is to be calculated from other values (in the same document
or not).
2) That the component is equivalent to a specific component in a canonical
schema.
3) That the value is to be read from some external data source.
4) The abstract role of a component (a la Dublin Core).
5) How the value is to be presented to the user visually.
6) ...

I suppose some will argue that these aren't really semantic characteristics.
Like most things, semantics is in the eye of the beholder. But if we can
automate tasks that formerly had to be performed by humans, there is some
semantic information being modelled, in my opinion. 

Matthew