- From: Jim Hendler <hendler@cs.umd.edu>
- Date: Mon, 26 May 2003 16:01:55 -0400
- To: Jeremy Carroll <jjc@hpl.hp.com>, webont <www-webont-wg@w3.org>
- Cc: "Peter F. Patel-Schneider" <pfps@research.bell-labs.com>
Couple of responses : A small quantity of beer was drunk at Jim's expense by Peter and myself on >Friday evening. A "small" quantity? Do I get change :-> (actually, if all the below flies, it was money well spent!) > >We considered the following 7 issues: [snipping] issues A - C look fine > >D) OWL DL syntax NP complete > >Included in comments >http://lists.w3.org/Archives/Public/public-webont-comments/2003May/0083.html >http://lists.w3.org/Archives/Public/public-webont-comments/2003May/0052.html >owlsas-rdfcore-np-complete > >Original message: >http://lists.w3.org/Archives/Public/www-webont-wg/2003Apr/0003.html > >We agreed on solution 2 at the end of the original message. > >change mapping rule to >EquivalentClasses( d1, .... dn ) >==> > T(dj) owl:equivalentClass T(di) . For all i,j s.t {i,j} in G, an arbitrary >connected graph over {1, 2, ... n } > >(Exact wording is editorial) >We agreed that the expected reader of S&AS is also expected to understand >terms from graph theory without any explanation, i.e. this rule will be >opaque to the non-specialist. I am not sure I understand the solution - does this make the mapping non-NP-complete by making it NP-hard or by moving it into a lower class :->? Seriously, I am not sure I understand why the solution above resolves the problem Jeremy brought up (the reduction of the mapping to a NP-complete graph problem) - I do know graph theory, but the above doesn't immediately make it clear why this fixed the problem. I can live with that, but writing a response to the rdf core group and others will require an explanation as to how this fixes the problem. > >E) unnamed individuals >http://lists.w3.org/Archives/Public/public-webont-comments/2003May/0052.html >owlsas-rdfcore-bnodes-restrictions > >Peter and I propose we postpone this issue. >Rationale: >- the amount of work to do otherwise is large, and is not guaranteed to >succeed. >- however the point is a good one and should be reviewed once the community >has more experience. > >(Personal note: my reading is that much of the approach taken in S&AS part 3 >would need substantial reworking to address this) > I think postponing makes sense - but I suspect we'll need something in our docs somewhere to talk about the implications -- hwoever, I think we're going to have to work harder to get all our docs to explain that RDF users should assume they will be using OWL Full unless they are willing to restrict their graphs - and then we need to explain the restrictions -- which probably means bringing that document of Jeremy's up to date or producing something else (see below) - Brian McB, Eric M. and a few others indicated to me in Budapest that doing this would help address their concerns- editors might want to think about thwre this explanation goes >F) ambiguity Just to confirm - you think the current solution is okay, but needs better explanation - and that would be localized to S&AS, is that right? > >G) OntologyProperty works for me >==== > >We also discussed the issue of an alternative articulation of the mapping >rules to go in the other direction. We were not convinced that these comments >could be addressed with a relatively small amount of work. A possible >approach is to sidestep them by pointing to two implementations of an OWL >Syntax Checker (which we already have, modulo the last details). This is the >initial proposal from the beer session. > >The possible additional text that we have is: >- the description of RDF graphs as triples >- the related proof (incomplete) > >The related proof does address the reverse mapping problem, but is unfinished >and difficult to read. It does include tables (which I don't think I ever got >round to generating) of when you see this triple then use this abstact syntax >construct. The amount of work needed to get it finished is nontrivial. > >The description of RDF graphs as triples could be updated failry quickly, but >only permits the building of a recognizer (an OWL Syntax Checker) and not a >tool which outputs the abstract syntax tree. How hard would it be to do a sort of "easy version" in the following sense - is there a subset of the OWL DL graphs that is easier to explain? In Judaism we call these things "normative" practice (note the very different use of the word normative, which confused me no end for a while) -- normative practice is a set of rules that oversimplify complex things in a way that is easier to follow. For example, the actual law in Judaism is something like "on the Sabbath you cannot write two or more letters next to each other with anything that either leaves a permanent mark whether it be ink or non-ink. If the letters are written at separate times, and they are within certain size limitations, then (this goes on for about 7 pages)" the normative version is "Don't write anything on Shabbat" What I'm thinking it would be possible to have a useful set of rules that says "if you live by the following rules, your graphs will be in OWL DL. There are other graphs which are also in DL, but are harder to explain in detail -- see S&AS for the full definition" guess I'm wondering of such a thing is possible... -- Professor James Hendler hendler@cs.umd.edu Director, Semantic Web and Agent Technologies 301-405-2696 Maryland Information and Network Dynamics Lab. 301-405-6707 (Fax) Univ of Maryland, College Park, MD 20742 240-731-3822 (Cell) http://www.cs.umd.edu/users/hendler
Received on Monday, 26 May 2003 16:02:04 UTC