- From: Birte Glimm <birte.glimm@comlab.ox.ac.uk>
- Date: Mon, 15 Feb 2010 20:45:23 +0000
- To: Axel Polleres <axel.polleres@deri.org>
- Cc: Lee Feigenbaum <lee@thefigtrees.net>, Ivan Herman <ivan@w3.org>, SPARQL Working Group <public-rdf-dawg@w3.org>
On 15 February 2010 16:29, Axel Polleres <axel.polleres@deri.org> wrote: > > Well, these (32 and 34) are definitly separate issues, but I guess we agree on that... Yes and 32 just is not related to entailment regimes. > As for issue-34, I agree that we can close the issue if the single goal is that the > behaviour is well defined... we have to be aware though, that results may not always > and for everybody will be intuitive: in fact, although every current implementation > probably treats bnodes just as constants under a kind of unique names assumption > in aggregates like count (does anybody do anything different?) there would > be alternative ways to handle bnodes in aggregates, like > Alternative1: ignoring bnodes for aggregates (could be viewed reasonable conceptually, IMO) That is an alternative, but I like it less. For example the OWL WG test results are recorded in an ontology that contains, for each test a bnode (anonymous individual in OWL) which is related via several properties to properties of the test in question, e.g., whether the test is a consistency or an entailment test, whether the test failed or passed etc. If you would ignore the bnodes/anonymous individuals, you couldn't count the number of tests or the number of failing tests etc. I can list it as an alternative, but I wouldn't vote for including this as the envisaged behavior. > Alternative2: for entaiment regimes like OWL which have an explicit notion of equality/inequality > counting could be expected to just return the number of those > - Alternative2a: known to be different > - Alternative2b: not known to be the same That means, however, that aggregation is no longer an algebra operator as it is at the moment. E.g., if we query a graph/ontology containing: ex:a a ex:C . ex:b a ex:C . (in FSS ClassAssertion(ex:C ex:a) and ClassAssertion(ex:C ex:b)) with a query SELECT ?x WHERE { ?x a ex:C } then we get x/ex:a and x/ex:b under OWL entailment (Direct Semantics and RDF-Based Semantics). Aggregates are then defined on top of that and just sum up or count or whatever the normal matching algorithm provides. I really wouldn't want to mess with that. Birte > Before just closing the issue, I would at least think that a sentence or two (and maybe examples) > should be added that clarify that those behaviors are NOT provided by the current semantics. > > Axel > > > On 11 Feb 2010, at 19:05, Birte Glimm wrote: > >> Just 32 means 34 I think: >> [ISSUE 34]: How do entailment regimes interaction with aggregates, >> grouping, and blank nodes? >> Birte >> >> On 11 February 2010 14:16, Lee Feigenbaum <lee@thefigtrees.net> wrote: >> > On 2/8/2010 2:02 AM, Ivan Herman wrote: >> >> >> >> I certainly agree with closing issues #28, #32, #40, #42. >> >> >> >> I would propose to leave #43 for a while to see how the SD work evolves. >> >> I predict that what you write below is true, but maybe discussions on SD >> >> will allow for a finer granularity... >> > >> > I will close issues 28, 32, 40, and 42 if there is no concerns raised within >> > the next 7 days. >> > >> > Let's include issue 43 in our next TC discussion of service description. >> > >> > Lee >> > >> > >> > >> >> >> >> -- >> Dr. Birte Glimm, Room 306 >> Computing Laboratory >> Parks Road >> Oxford >> OX1 3QD >> United Kingdom >> +44 (0)1865 283529 >> >> > > > -- Dr. Birte Glimm, Room 306 Computing Laboratory Parks Road Oxford OX1 3QD United Kingdom +44 (0)1865 283529
Received on Monday, 15 February 2010 20:45:56 UTC