- From: Sankar Virdhagriswaran <sv@crystaliz.com>
- Date: Sat, 20 Nov 1999 21:43:18 -0500
- To: Dan Brickley <danbri@w3.org>, www-rdf-interest@w3.org
Just to make sure: I am really not pushing for 'the' entity-relationship model. What I am pushing for is the simplest model that we can get away with at the core of RDF spec. Since RDF at the core is a variation on 'the' entity-relationship model. Why not start proselytizing and using that model. >One perspective is that RDF's terminology and specification style >obscures the similiary between RDF and the ER model. > Yup. Furthermore, we have added a lot to it. >and resources and predicate/subject/object and properties and statements and >rdf descriptions, we risk obscuring the essential simplicity: the RDF >information model is just the Web information model. On the Web, I could not have said this better. >differences) between RDF and (amongst other things) E-R. > Furthermore, while to folks coming from the inferencing/logic camp predicate/subject/object may look like a simple addition, it is actually conceptually very hard for anybody who does not have that training. Again, there is past history to help us here. ER models have been successfully used for a long, long time in the commercial world. However, Datalog which uses very limited predicate calculus has not found use in the commercial sector. As anybody who has played with both know, Datalog can express a lot of relationships very easily and can also perform powerful queries in a very scalable fashion. However, due to the presence of SQL, what has happened is that Datalog concepts have been folded into SQL 3. We have a similar problem here. We have the XML-Schema folks defining things that will support SQL like queries. And, I predict that the XML Query language group will follow and define a query language that will look like SQL. I also predict that once that happens, we will again have these arguments with a "Charles River Communiqué" which will say nothing and create more confusions. Also, a general relationship model can be easily extended to support predicate/subject/object model. So, why not start with some thing that is brain damaged simple and layer more and more complexity on top. Sankar
Received on Saturday, 20 November 1999 21:43:56 UTC