- From: Hassan Aït-Kaci <hak@ilog.com>
- Date: Mon, 08 May 2006 15:23:05 -0700
- To: Gary Hallmark <gary.hallmark@oracle.com>
- CC: W3C RIF WG <public-rif-wg@w3.org>
Gary Hallmark wrote: > > I agree that there are many ways to approach semantics. While I'm not > sure I would characterize model-theoretic semantics as more or less > abstract than other approaches, I do fear that a model-theoretic > semantics will be of little help > to the implementors of RIF translators and associated rule engines. > Looking at other W3C formal semantic specifications for guidance, I find > http://www.w3.org/TR/xquery-semantics/ to be a good approach. Formal > XQuery semantics are specified using RULES. Why can't we specify RIF > semantics using rules? We could even write those rules using RIF. Amen to this! At last! That was my point precidely as expounded (perhaps too early in the process?) in the following RIF mail archive messages: http://lists.w3.org/Archives/Public/public-rif-wg/2005Dec/0096.html http://lists.w3.org/Archives/Public/public-rif-wg/2005Dec/0083.html http://lists.w3.org/Archives/Public/public-rif-wg/2005Dec/0081.html In passing, the XQuery semantics is precisely using Gilles Kahn's Natural Semantics (NS) (NS = Plotkin's SOS used with Horn Rules expressing Gentzen Sequents) - I gave the references in the mails above. Such rules are used universally by most Programming Language formalists (see any POPL paper) to express formally static (i.e., compile-time) and dynamic (i.e., run-time) semantics of languages. XQuery's semantics was the work of essentially Phil Wadler whose Programminhg Language background accounts for the use of Natural Semantics style of formal semantics. The advantage using NS kind of rules expressed in RIF to specify semantics of a RIF-compliant language would even make possible a just-in-time interpretation of constructs requiring a semantics not handled by a client. Indeed, instead of ignoring the requirement or failing, such a client could use the RIF-specified semantics to interpret the constructs whose semantics it does not support. This could be done by using a universal NS rule interpretor (e.g., a URI offering NS rule interpretation services for rule sets expressed in the RIF). In this manner, the RIF semantics would be a meta-semantics expressing the formal semantics of families of languages in a way to make them operational. The beauty of this is that - a Gary has noted - it expresses the semantics of rule languages using a rule language. -hak -- Hassan Aït-Kaci ILOG, Inc. - Product Division R&D tel/fax: +1 (604) 930-5603 - email: hak @ ilog . com
Received on Monday, 8 May 2006 22:23:20 UTC