See also: IRC log
- Doug Lenat, Sandro Hawke, Rinke Hoekstra, Alan Ruttenberg, Jeremy Carroll, Michael Smith, Ratnesh Sahay, Ivan Herman, Bernardo Cuenca Grau, Markus Krötzsch, Zhe Wu, Ian Horrocks, Evan Wallace, Peter Patel-Schneider, Jeff Pan, Uli Sattler, Deborah McGuinness, Boris Motik
- Martin Dzbor, Achille Fokoue, Jim Hendler, Joanne Luciano, Bijan Parsia, Michael Schneider
- Alan Ruttenberg
- Doug Lenat
(Scribe changed to Doug Lenat)
Alan Ruttenberg: accept minutes from Jan 9
PROPOSED: accept minutes of Jan 9
RESOLVED: accept minutes of Jan 9
PROPOSED: accept minutes of last week
RESOLVED: accept minutes of last week
Discussion (quick) of pending actions
Proposals to Resolve Issues
Issue 8: as proposed (not clear how to do this, hence postpone)
PROPOSED: close (as POSTPONED) Issue 8 (dataproperty chains) as per http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0122.html
RESOLVED: close (as POSTPONED) Issue 8 (dataproperty chains) as per http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0122.html
Turning to Issue 15 (Ontologies should not be required to include a URI) as per http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0019.html
PROPOSED: close (as RESOLVED) Issue 15 (Ontologies should not be required to include a URI) as per http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0019.html
Deborah McGuinness: We'd be better off with named ontologies, and should indicate it's a "best practice"; easier to track where they came from.
Jeremy Carroll: concern about further diversion from owl full and its family members.
Peter Patel-Schneider: we don't have to right now, we allow unnamed ones, so there is no need for serious concern about that
Alan Ruttenberg: we want imports, want synergies, so why not encourage it?
RESOLVED: close (as RESOLVED) Issue 15 (Ontologies should not be required to include a URI) as per http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0019.html
PROPOSED: close (as RESOLVED) Issue 29 (User-defined Datatypes: owl:DataRange vs rdfs:Datatype) as per http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0017.html (see also http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0147.html and thread)
Peter Patel-Schneider: summarizing the issue: proposal to make owl datarange unnec. except for backwards compatibility
RESOLVED: close (as RESOLVED) Issue 29 (ser-defined Datatypes: owl:DataRange vs rdfs:Datatype) as per http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0017.html (see also http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0147.html and thread)
General Discussion (Punning)
Punning discussion beginning now. Note that this will not be leading to a resolution today, this is intended only to be a discussion. Please see http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0231.html for a list of the 5 questions about punning that we will be referring to, during this discussion.
First we consider question 1: Consider the punning issue to be divided into two kinds of punning. The first kind adds instance punning against classes and properties. In some sense this is the most easily understood kind of punning and those for which there are obvious use cases. The second kind are the other punning pairs - class/property, objectproperty/ dataproperty. Is it worth considering these separately? Do we have any kind of consensus that one or both are desirable/useful?
Doug Lenat: It seems that there are no clear opinions on whether to distinguish those 2 kinds of punning.
Turning now to question 2: Two cited cases for punning are Metamodeling and being able to have real properties on classes/properties. But what exactly do people consider Metamodeling, and does the punning proposal address these cases. As an example, it does not address the cases on Conrad's Metamodeling page because we don't plan to support modification of owl syntax.
Boris Motik: "services" are instances, take instances of people,... the value of the property is not a particular person but the class Person.
...Monkeys eat bananas meaning that EACH instance (each individual monkey) does this, and Monkeys is itself an instance of Species(Type)
...there is a reasonable case for treating classes as instances and properties as instances
Markus Krötzsch: the use case that Boris mentioned is similar to my experience. City instances have the property population, e.g.
...(Markus continues:) You don't know the types of things you have in advance, at least not always.
Ivan Herman: Dublin Core had to worry about this with e.g. different ways to identify a person
Boris Motik: in these examples, people are just stating things, and using idiosyncratic code to reason with those assertions.
Alan Ruttenberg: people want to restrict domains and ranges on class properties
...punning lets us use real properties and subproperties this way.
...Can someone compare data and object property punning to the use of annotation properties (pro or con)?
Boris Motik: when it comes to annotation-spaces, they require this type of punning.
...given an annotation in the original ontology, you need punning because in the new ontology it might be one of two or more types.
Alan Ruttenberg: what about the annotation properties in Owl 1.0 -- still okay? still there?
...[Alan continues:] in what sense is that not enough, why do we need to use punning for that now, suddenly?
Jeremy Carroll: let's record these various use cases (on the wiki)
ACTION: Markus to put his use case for punning onto a wiki page
ACTION: Boris to put his use case for punning onto a wiki page
ACTION: Alan to put his use case for punning onto a wiki page (for reasoning associated with punned properties, and the question about object data property punning vs annotation properties)
Boris Motik: annotation properties (in OWL DL) not real properties
...they can have both indiv and data values
Evan Wallace: UML has association class, a natural way to talk about properties of relations.
...not its defintion, properties of the relation. Commonly used in UML.
ACTION: Evan action to describe UML association class on the wiki, w.r.t. punning discussions
Alan Ruttenberg: use cases where people want to do logical inference not just recording the assertions, please.
Michael Smith: describing another case, may be the same as Markus' essentially.
Turning now to the next question, number 3: From a technical point of view, how would dropping some or all of punning help? To what extent is the amount of new vocabulary dependent on our choice of punning? How does punning effect OWL Full?
Peter Patel-Schneider: I repeat, the answer to question 3 is NO.
...so now we have to come up with three fixes for these three new problems.
...in order to allow punning.
...the owl DL and owl FULL gap is widened more than it would otherwise be, by allowing punning.
Peter Patel-Schneider: disagree: if you're not limiting yourself to OWL DL, why should you care?
... Complicated mapping rules are worth it, if it makes the conversion of triples into ontology easier.
Boris Motik: why not allow a few more properties here and there, what is bad about that?
Alan Ruttenberg: tried to separate 2 issues. Incr. vocab makes the parsing of triples easier and deterministic.
...the addl vocab removes ambiguities and localizes information.
...This is a nice-to-have but not a need-to-have feature.
...But allowing obj/data polymorphism makes it more of a requirement than an option.
Peter Patel-Schneider: you may need some fancy footwork, to avoid ambiguity.
Jeremy Carroll: to answer Boris: larger vocabularies means more work and may actually drive off other potential OWL users.
Boris Motik: punning doesn't necessitate more vocab actually.
...(Boris continues:) using typing to specify the type of partic properties.
...(Boris continues:) if there is no obj/data punning.
Peter Patel-Schneider: using rdf + a litle bit of owl means what, exactly? you're probably not in owl-dl any longer.
Moving on to question 4 of http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0231.html Namely: From a communication/understandability/documentation point of view, how would our choices effect the communities that we want to use OWL. What is the extra documentation needed to explain punning? How much would eliminating or reducing punning help? What's the appropriate balance of cost/benefit?
Ivan Herman: agree with Jeremy; I'm afraid of bloating the number of things a user needs to learn.
...most users use some editing tool; for them, learning more is a major issue.
Uli Sattler: looking to the future, fewer and fewer people will be using standard editors to
...manually write OWL ontologies. Future tools will take care of more the bookkeeping details.
Peter Patel-Schneider: agree.
Ivan Herman: there will be users who create complex ontologies and will need future sophis. tools.
...but folksonomies staying at the DLP level is another more common situation.
...and those users will be around for many years, using EMACS eg and that's it.
...they won't be using Protege and similar tools that go way behond what they need.
Deborah McGuinness: stepping back a bit, to the general issue of adding more and more new constructs...
...maybe there is such a thing as too many.
...TO handle that, maybe select fragments ahead of time to help them focus on a useful subset or two.
Doug Lenat: good point, Ian.
Decide whether to discuss these next N issues, or skip them. For a listing of the N issues, see Raised Issues in http://www.w3.org/2007/OWL/wiki/Teleconference.2008.01.23/Agenda
No one minds discussing issue 95.
Issue 94 Problem with roundtripping when going from functional-style syntax into RDF and back
Doug Lenat: -1 (because it's not completely clear that we want to support such roundtripping; we should explicitly discuss why or why not to provide such a guarantee.)
Doug Lenat: Enough people want to discuss it, so yes, let's discuss it.
Alan Ruttenberg: think also about whether and to what extent round-tripping is a requirement.
Issue 93 RFC 3066 - Tags for the Identification of Languages
Backward compatibility nits: see Issue 90, Issue 91 and Issue 92
...Options related to deprecating classes, properties and datatypes
Peter Patel-Schneider: let's not preserve deprecation, though versioning is still required.
Alan Ruttenberg: should we ignore them, treat them as annotations,...?
Peter Patel-Schneider: depr. is special; it's wierd,... may be best strategy to just ignore it and see if something bad happens.
Jeremy Carroll: if we do deprecate deprecation, then we better ask other groups whether there is some unintended bad synergy with what they are doing.
ACTION: jeremy to edit references in syntax to address addison's bcp47 comment
ACTION: pfps (Peter Patel-Schneider) to write proposal to resolve 90 to drop deprecation and document in changes
ACTION: Patel-Schneider to write proposal to resolve 90 to drop deprecation and document in changes
Turning now to http://www.w3.org/2007/OWL/tracker/issues/91
Rinke Hoekstra: summarizing what it says on that link.
Doug Lenat: Note that "ontology properties" are as follows: owl:imports, owl:priorVersion, owl:backwardCompatibleWith, and owl:incompatibleWith.
Peter Patel-Schneider: the state of owl:imports is: directive
Alan Ruttenberg: leave it as it is, not an ontol. property; make it subservient to versioning.
Turning to Issue 92: The RDF to FS mapping does not provide a mapping for the owl:Ontology element to the Functional Style syntax
PROPOSED: it's just a bug, so let's just fix it (i.e., treat it as nothing more than an editorial action).
Doug Lenat: So the consensus is to treat it as an editorial issue (not something to discuss)