Re: RDF/S entailment rules.

Hi Pat,

I see your point but, sorry about that, I do not agree with your conclusion...

- The papers you refer to are copyrighted to Elsevier and Springer, respectively. Your first reference is unaccessible to most; e.g., because I am currently at home and not at my institute, I cannot download a copy from Elsevier's science direct without paying €40.-. The second paper is available but, formally, Herman is breaking copyright law by putting it out there through his web site; Philips may decide to pull it down from the Web because of that (or if Herman retires or leaves Philips). Bottom line is that we have no security that those results will be available to anyone in a few years' time for free. That is against the, shall we say, 'ethos' of royalty free W3C documents.

- I have read those papers and yes, I have actually used them to implement RDFS in a straightforward way. And they are great papers, absolutely no doubt about that. But... it was not an easy read. Herman is a mathematician, writes as a mathematician, and for mathematicians; beyond the style, the paper also includes a number of issues, theorems, etc, that an implementers are not necessarily interested in. The paper also includes rules sets for something that was a precursor of OWL 2 RL (pD*) which goes way beyond what we need. Meaning that, even for an implementer, reading through the paper and find what he/she really needs as way tougher than necessary.

- I believe that the usefulness of the documented rule set is not only for implementers. The rule set, in my view, helps the everyday user in understanding what is going on in general, it helps to establish some sort of a mental model of what RDF(S) entailment does for you. Even if the rule set is incomplete or not 100% precise, it is still immensely useful for most of the users. Those users will certainly not read Herman's paper, they will be scared away by the abstract or the first section... and we leave them with nothing.

Consequently, I believe we should document something in W3C space. I would perfectly o.k. deferring the mathematical theorems on completeness, complexity, etc, to an external paper, ie, to Herman's papers. But I would still prefer to see the basic rule set reproduced (with due reference to Herman, of course) in our documents, even if the only thing we do is to create a set of tables like the ones used in the OWL 2 RL spec[1].

Cheers

Ivan


[1] http://www.w3.org/TR/owl2-profiles/#Reasoning_in_OWL_2_RL_and_RDF_Graphs_using_Rules




On May 5, 2013, at 17:10 , Pat Hayes <phayes@ihmc.us> wrote:

> I have a kind of editorial progress report to give. Or maybe lack of progress.
> 
> The 2004 Semantics document included a long section describing sets of entailment rules for various entailment regimes, with completeness theorems plus proofs, which I suspect only two people ever read. The rules for RDFS were actually incomplete, and the proofs were wrong. None of this was normative and none of it was needed to expalin the actual semantics. During our WG discussions I suggested simply removing it from the document, but this idea met with some pushback from members who felt that these rules had been the most accessible way for some readers to understand the semantics. One idea that was suggested was to make the rules section into a separate document. 
> 
> I have been drafting this separate document, until I realized that it has already been written, by Herman ter Horst. Herman analyzed the rules in the 2004 document, found the subtle errors in them, invented an elegant improvement of them which has been fully described in a series of published papers, and even implemented in a publicly available Python app. He also proved completeness and established some decision and complexity results. Anything I write will be little more than a direct copy of his work, with footnotes. 
> 
> I think the only sensible thing is to refer to this published work, and invite interested readers to check it out and make the small adaptations that might be necessary to handle the small changes in the semantics since 2004. Anyone who is interested enough to actually implement this stuff will be able to do this. 
> 
> Comments?
> 
> Pat
> 
> Herman's papers:
> http://www.sciencedirect.com/science/article/pii/S1570826805000144
> http://www.extra.research.philips.com/hera/people/terhorst/ISWC2005p668.pdf
> 
> ------------------------------------------------------------
> IHMC                                     (850)434 8903 or (650)494 3973   
> 40 South Alcaniz St.           (850)202 4416   office
> Pensacola                            (850)202 4440   fax
> FL 32502                              (850)291 0667   mobile
> phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes
> 
> 
> 
> 
> 
> 


----
Ivan Herman, W3C Semantic Web Activity Lead
Home: http://www.w3.org/People/Ivan/
mobile: +31-641044153
FOAF: http://www.ivan-herman.net/foaf.rdf

Received on Monday, 6 May 2013 11:47:54 UTC