Re: [RIF][UCR] Decidability a requirement?

Christian de Sainte Marie wrote:
> 
> [...]  we encourage you to
> propose, on this medium, arguments in favor or against making 
> decidability a requirement for all or part of the RIF.

[Chair's hat off]

Decidability, in the sense that "RIF allows only the expression of rules 
or rulesets that are decidable wrt any RIF expression", has nothing to 
do with requirements or design constraints for the RIF, IMHO (or: didn't 
we decide that it was "le RIF"?).

If a rule language covered by the RIF is decidable (in the same sense, 
mutatis mutandis), then it will be able to produce or consume only 
decidable RIF rulesets (decidable wrt expressions in that same rule 
language). If all the rule languages covered by the RIF are decidable, 
then only decidable rulesets will ever be actually expressed in the RIF 
(used as a RIF; and decidable wrt expressions in the rule languages that 
produce/consume each of these rulesets, resp.).

None of the above requires that the RIF itself should be decidable.

Of course, the requirement in question could be that the "RIF is 
required to cover only decidable rule languages". But, (i) if somebody 
somewhere needs to exchange non-decidable rulesets bewteen non-decidable 
rule languages, I do not see why they should be restricted to use the 
RIF for that purpose; and (ii) if we are talking of RIF Core, I do not 
see that this be a particularly useful restriction (the reason for 
restricting RIF Core expressiveness being to focus our attention on the 
foundation layer of the RIF, that is, on the architecture and principles 
(extensibility, conformance, the XML syntax, compatibility and 
interoperation with relevant standards, esp. RDF and OWL, data access 
etc), with important but difficult features being postponed to the 
Extended RIF - and the charter already mentions "essentially Horn logic" 
for that purpose).

[Chair's hat on]

CHristian

Received on Thursday, 6 April 2006 16:34:53 UTC