- From: Graham Klyne <gk@ninebynine.org>
- Date: Tue, 04 Nov 2003 16:56:02 +0000
- To: "Jeremy Carroll" <jjc@hplb.hpl.hp.com>, <www-rdf-logic@w3.org>
Hi Jeremy, Thanks for commenting my thoughts -- that was useful confirmation of some of my ideas. At 16:07 04/11/03 +0100, Jeremy Carroll wrote: >Doing fancy stuff in restrictions is very hard. Well, maybe. What I'm trying to do isn't particularly fancy. The restrictions approach seems to work nicely within the existing RDF syntactic framework. Also, it seems to provide a more uniform approach to expressing application domain knowledge. If I know a+b=c, then given any two of a,b,c I can deduce the third. Using conventional rules, that takes three separate rules. When I was working on a network device configuration application using RDF earlier this year, I found that I typically needed only very simple patterns of inference. "A little inference goes a logn way". What was missing from the tools I was trying to use was the ability to handle odd datatypes (notably IP addresses). So I'm looking for ways to encode knowledge of datatypes in a way that applications can easily access. I've just settled on trying an approach based on class restrictions. I think it's worth experimenting with. >Without any named variables life is tough. >Rules languages tend to add variables. At some level, I'm sure you're right. But I glimpse parallels with the functional programming I've been playing with, and in particular a "point free" style of programming that avoids the use of variables. It takes some getting used to, but often results in very elegant expressions. #g ------------ Graham Klyne For email: http://www.ninebynine.org/#Contact
Received on Tuesday, 4 November 2003 12:26:11 UTC