W3C home > Mailing lists > Public > semantic-web@w3.org > March 2005

RE: true/false in RDF?

From: Joshua Allen <joshuaa@microsoft.com>
Date: Thu, 17 Mar 2005 07:54:52 -0800
Message-ID: <0E36FD96D96FCA4AA8E8F2D199320E520488400F@RED-MSG-43.redmond.corp.microsoft.com>
To: "Geoff Chappell" <geoff@sover.net>, "Danny Ayers" <danny.ayers@gmail.com>
Cc: "Seth Russell" <russell.seth@gmail.com>, <semantic-web@w3.org>

> Or do you think that it's the knowledge overhead/learning curve of
> rdfs/owl that limit its use to a subset of the developer population?

Yes, exactly.  We followed the same approach with XSD and SOAP/WSDL --
the theory was that people would define their interop in a data-driven
manner, and that would make it easier for anyone and everyone to
interop.  We have five or so years experience doing this with XML.  For
example, given an XSD, I can autogenerate Java or C# classes to parse it
with full typing.  Or, given a WSDL, someone using J2EE can write proxy
calls to a web service on .NET.

But what we see in practice is that each extra level of contract chops
off a layer of potential partners.  The XML formats which have the most
developers do not see those developers using XSD to figure out how to
code against the format.  In fact, many do not even use namespaces.
Developers copy/paste code from books, or look at some sample XML (or
English prose documentation) to figure out how to write their code.  In
other words, the percentage of people who use the schema as an authority
on the structure of their XML for programming purposes is a small
fraction.

Another example, BPEL4WS is a data format which defines workflow, and we
have proven interop between all the large ISVs.  But in reality, the
number of people who can successfully use these ISVs' products to
interop via BPEL4WS is way smaller than the number who can use SOAP.
And the number who can successfully use SOAP stacks is much smaller than
the number who can integrate with paypal using HTTP redirect and
postback.

So the argument always was that these schema layers enable *broader*
interop.  But at the present time, I disagree.  They enable *better*
interop, among deeply-integrated enterprises, but they actually *shrink*
the overall scope of interop.  I also think it is a false hope to think
that tools will solve the interop problem.  Interop with no tools
required always trumps a good tool.

So if we see semantic web being used to do really wonderful things in
the scope of enterprises the way SOAP is being used to deploy web
services in the enterprise, then I embrace the whole RDFS/OWL stack.
But if we want wide deployment like flickr, del.icio.us, and RSS -- like
the original WWW that is to say -- then we need to focus on the part of
the stack that mere mortals can use.  The nice thing about OWL is that
it doesn't break with the simple mortal subset of RDF that I'm
proposing.
Received on Thursday, 17 March 2005 15:55:29 UTC

This archive was generated by hypermail 2.3.1 : Tuesday, 1 March 2016 07:41:45 UTC