RE: Web services and the Semantic Web

> -----Original Message-----
> From: Sanjiva Weerawarana [mailto:sanjiva@watson.ibm.com]
> Sent: Sunday, May 26, 2002 1:24 AM
> To: Mark Baker; Eric Newcomer
> Cc: www-ws-arch@w3.org
> Subject: Re: Web services and the Semantic Web
> 
>
> 
> I read this, but I'm still left with the same "didn't grok it"
> feeling. One of your statements struck me in particular:
> 
>     "The neat part of RDF and the Semantic Web is that I can
>      build a purchasing agent that knows nothing about shoes."

Hmmm, using a universal Turing machine simulator, I could write software
that knows nothing about computers :~)   As the aphorism goes, "there is no
problem in computer science that can't be solved by adding an additional
layer of abstraction."  It's far less clear that these beautiful
abstractions help much in the real world of commercial software ... see the
famous "Worse is Better" essay.
 
>  If I undertand the
> scenario correctly its something like this: there's a program 
> ordering someone's wardrobe. It will somehow infer that that 
> that means they need to order shoes. 

I think that's where "ontologies" come in. There would be some network of 
assertions that knows that a "wardrobe" (in some culture, for some gender,
for some age, climate, etc.) consists of shoes, socks, trousers, shirt, etc.
I suppose that some other "ontology" would understand the process of
ordering clothing (determine size, find a supplier, negotiate a price, place
an order ...).  The point, I guess, is that all these networks of assertions
can be processed by quite generic code -- rather than writing procedural
"buy a wardrobe" or "shop for shoes" code when a need for it arises, one
builds RDF data (linked to other ontologies on the Web).

More realistically, I would guess that the ontologies that one would develop
in the real world web services context would be at a much lower level.  For
example, presumably WSDL could be replaced by an ontology for describing web
services invocation and responses, and a security markup language replaced
by an ontology for describing authentication, access control, encryption,
etc.

> Web services are really just standardizing lots of existing
> practice. People already do things that SOAP standardizes, 
> have ad hoc ways to solve the WSDL problem, and so on. By
> standarding a base layer, we're enabling the next layer of
> problem to be solved, including the scenarios offered by the
> semantic Web.

I agree.  I have a somewhat schizophrenic attitude toward this
stuff ... I would like to believe it will work, but it sounds
too much like AI things that I was excited about in grad school 
25 years ago (sigh) that were going to be big ... Real Soon Now.
Still, the thing I most clearly remember from my AI class
in 1977 is probably still true today -- this stuff works best
in quite limited domains (or what a colleague calls "the really
banal level" 
http://lists.w3.org/Archives/Public/www-rdf-interest/2002Feb/0045.html )

Maybe ontologies for web services description or security assertions
or reliable message transmission are sufficiently "banal" to be
practical.  I'll be happy to see this demonstrated with simple
specs and running code ... but I'm too jaded and cynical to get excited
once again about something that is gonna be big ... Real Soon Now.

Received on Sunday, 26 May 2002 09:47:57 UTC