- From: Sankar Virdhagriswaran <sv@crystaliz.com>
- Date: Wed, 17 Nov 1999 08:37:00 -0500
- To: Stefan Decker <stefan@db.stanford.edu>, www-rdf-interest@w3.org
If I had my choice I would propose the following. A simple RDF will have only the basic capability to model entity-relationships. One can go back to Chen's original paper and take what he proposed and implement that in RDF. No more, no less. No logical expressions, no constraints. There are lot of tools out there that can create entity-relationship models and more importantly they are all wired up to things that are real (RDBMSs, document databases, etc., etc.). So, it is an easy step for any one using those tools to export their conceptual schema to the Web if they choose to. Just that will go a long way in writing 'agents' that understand entity-relationship models do intelligent things for the user. We will also avoid the10th rule problem found in inferencing systems of the past (adding the 10th rule sends the inference engine for a loop and no body knows why -- this in a centralized development environment). A slightly more sophisticated implementation of this idea is Sergey's proposal on relationships. Let us start there. I think the problem is not just simple 'syntax'. It is simple conceptual implementation. Anytime we want to add things to help in inferencing (such as the Not proposal), we need to think through whether it will scale from the perspective of the consumers (i.e., inferencing agents) and whether novice producers will understand such expressions or the impact of those expressions on inferencing. Sankar
Received on Wednesday, 17 November 1999 08:35:14 UTC