Re: [xml-dev] XML vocabulary for expressing constraints?

[Steve Newcomb <srn@coolheads.com>]

On 12/11/2013 06:31 AM, Hans-Juergen Rennau wrote:
...
> The change of paradigm I am waiting for is a new perception which sees
> documents as contributions to a single, homogeneous space of
> information, I am waiting for a widespread recognition that documents
> have a dual nature - distinct entities, *and* just part of a homogeneous
> substrate of information, which is the forest of information items
> effectively created by the sum total of accessible XML resources.

Unless I misunderstand you, Hans, SGML has always had this.  As far as I
can see, XML still has most of it, amid a storm of deprecation by those
who don't accept the fundamental distinction between "a homogeneous
space of [all?] information" and a SYSTEM that provides access to some
portion of that space.  (Although it's relevant, I omit mentioning the
argument about NOTATION attributes; I don't feel like going there today.)

In my view, what SGML never had, and XML still doesn't have, is a
mechanism for disclosing the semantic intent of a document address.
When I mention an address, what am I talking about?  In the absence of a
characterization of intent, it's extremely ambiguous.

The whole question of intent can be evaded, sort of, if one implicitly
or explicitly denies that there's any (supported or supportable) intent
other than to provide access to the addressed component of the "space",
and only within the context of some system.  In other words, the
available system behaviors delimit the semantic possibilities.  When you
say "homogeneous substrate" I'm tempted to think that this is your
position: that everything is XML, and that the semantic that a node
derives from its context is simply that it *is* in its context.  It's an
honorable position; it has the virtue of internal consistency.  Much can
be accomplished within that scope.  But I don't think it's a general
solution to the problem of information interchange.

Another way of avoiding the question of how to disclose the intent of an
address is to muddy the waters by saying that all addresses, and
addresses that aren't resolvable in the context of any known system,
can carry *any* semantic load, as long as they look like addresses.  I'm
very uncomfortable with that approach.

I hope someday we will do better with disclosing semantics than we have
done.  The Bitcoin phenomenon gives me some hope, but I can't claim to
have solved the puzzle.  Far from it.  But I think (or, really, hope)
it's important to orient some thinking in that direction.