Re: GRDDL spec ready for release? (informative mechanical rules)

While GRDDL is seen as a "lightweight" (i.e. should be easy to read and
implement) spec, a bit of formality can be useful.

I agree with DanC's idea to put the N3 Rules out in a separate file, and
so make the spec look a bit less scary to people that don't know N3
Rules - but they would still be useful to implementers that know or are
interested in N3 rules.

I think Dan's normative formal semantics as given by his text is good
though, and I would prefer personally if it stayed in the spec.

             -harry





Dan Connolly wrote:
> On Fri, 2007-02-02 at 13:22 -0500, Chimezie Ogbuji wrote:
>   
>>> Thanks so much for all the hard work on the Spec! I'd like at least one
>>> more reviewer besides Ron to give a good read before we release it as a
>>> Last Call.
>>>       
>> I'll volunteer to help review the spec.  I'll try to focus on the 
>> informative mechanical rules, ...
>>     
>
> Any progress?
>
> On my side, I'm thinking about splitting the "mechanical rules" appendix
> to a separate file. Right now I have the first 34 steps of a proof
> sorta cleaned up; maybe we can walk thru it in tomorrow's telcon...
> or maybe some other time...
>   http://www.w3.org/2004/01/rdxh/spec_rules.html
>   revision 1.1 date: 2007/02/07 00:54:09
>
>   
>>  as I have a concern that there isn't much 
>> precedent in using rules to express the processing mechanics of a 
>> specification
>>     
>
> Sure there is... relaxNG, XQuery formal semantics, just to
> name a couple off the top of my head.
>
>   
>>  as well as the informal dependencies on the vocabularies 
>> used in the rules (the log:* properties in particular as well as 
>> rdfsyn:*).
>>
>> The appendix to the mechanical rules has @@explain TODO's 
>> regarding the vocabularies which are not 'formal', and I think these need 
>> to be very explicit about what the semantics of these terms are.
>>     
>
> I'd like them to be reasonably clear, but I wonder what "need" you see.
>
>   
>>   The fact 
>> that the rules rely on generating function symbols makes the underlying KR 
>> quite expressive and perhaps not aligned with the target KR of the 'sanctioned'
>> semantic web rule language (RIF) - which is still a work in progress.
>>
>> log:uri and log:includes in particular are quite cryptic in their formal 
>> semantics - at least from what I know of what those terms are meant to 
>> mean.
>>     
>
> log:uri is just like 'name' from KIF, and
> log:includes is just rdf simple entailment.
>
>   
>> Though the rules are informative, they have the nice advantage that they 
>> are 'complete' in the sense that proofs can be generated to determine 
>> GRDDL compliance.  The bar for the average GRDDL implementor becomes 
>> significantly higher if they intend to interpret the rules in any formal 
>> way, and I think at the very least we should be sure that the specificatin 
>> is as clear as it can be (given the fact that we still don't have a 
>> 'sanctioned' SW rule language) about such an interpretation.
>>     
>
> By all means, send any clarifications you can think of.
>
>   


-- 
		-harry

Harry Halpin,  University of Edinburgh 
http://www.ibiblio.org/hhalpin 6B522426

Received on Wednesday, 7 February 2007 01:34:52 UTC