W3C home > Mailing lists > Public > www-rdf-interest@w3.org > December 1999

Re: [www-rdf-interest] <none>

From: Sankar Virdhagriswaran <sv@crystaliz.com>
Date: Thu, 23 Dec 1999 21:42:51 -0500
Message-ID: <002b01bf4db9$5e986f40$e6ea7392@honeybee>
To: "Sean Luke" <seanl@cs.umd.edu>, <www-rdf-interest@w3.org>
Cc: <jhendler@darpa.mil>, <heflin@cs.umd.edu>
Sean,

> So why not just dump the special-case stuff and go with simple
> (simplistic) general-purpose inferential semantics to begin with?

I am wondering about two different types of scalability:

a) Agent writer(s) scalability: The 'real world' still is hacking away in
VBScript/Perl. For them, inferencing (of the logic/prolog kind) is too
foreign. Given an object model, they are more prone to write 'procedural'
agents. At best, we can hope them to write SQL queries. This was also the
case in the 80's when folks were writing expert systems. We had a tough time
convincing people to switch over to anything close to inferencing (even the
OPS 5, simple production rules kind). This was within corporations which
would make money if they did switch over. We are hoping that RDF will be
adopted by the world. There, IMHO, most users, at best are procedural agent
writers, not inferencing folks.

b) As Steven Decker points out there are scalability issues with predicate
calculus based inferencing. Moreover, only the top end ORDBMS products
implement most of what Datalog (which is arguably simple in terms of its
inferencing capability) can do. So, if one is interested in implementing a
'semantic search engine' for the web which will be described a 'semantically
clean RDF that allows inferencing', then one is forced to thinking about
building a beast that even the most seasoned engineering teams have
difficulty with. So, 'general-purpose inferential semantics', IMHO, has
created a big problem for scalability (i.e., network effect) for that idea.
Imagine where the Web would be today without search engines. Can you imagine
writing an 'inferencing engine' for the semantic web that can scale the same
way if one adopts your model?

I think we need to start very, very simple since we are looking at a
scalability that has not been attempted by any of the knowledge
representation projects I know. We are operating in a completely different
world as compared to the closed world situations where knowledge
representation techniques have been used in the past.

Sankar
Received on Thursday, 23 December 1999 21:48:31 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Monday, 7 December 2009 10:51:42 GMT