Lists Home |
Date Index |
It indicates the old saw, "The devil is in the details."
Don't bet the farm on a handshake then walk away.
XML is capable of enabling you to express meaning, data
and code, if you like. The infoSet can't. As soon
as you create and application language, you step beyond
XML and into the world of semantics. You can
always build an object framework to interpret that or
a compiler to build an executable image for some
platform. There is a reason for that PublicID
and it is deliberately separated from the SystemID
because they DON'T mean the same thing. The Public
agrees; the System is the local implementation of that.
The first is a contract. The second is a locator.
Binding those together is the central flaw in the
design of the web. URL <> URN <> URI of necessity.
It is convenient. Because of this, the web has to
go to the Semantic Web, but it is a Golem at that
point and you should both fear and respect that.
Agreements are often layered. In the original thread,
I asked if one could negotiate out the noise. In effect,
that is all we ever do. A semantic is declared and
agreed to (the ontological commitment - see Tom Gruber),
then the conformance to that agreement is tested and a
negotiation started if it is non-conforming. All
the system has to do is ensure that a set of choices
is reproducible at both ends of a communication (or
in the case of multicast, multiple endpoints).
From: Nicolas LEHUEN [mailto:email@example.com]
But we should think about the possibility of having assumptions at a higher
level, e.g. both sides agreeing on the concept of 'neural network' or
'classification engine' (domain specific), 'rendering' or 'validation'
(generic). As those concepts are expressed both in code and data, it means
looking further than XML, which concentrates on data.