Warning:
This wiki has been archived and is now read-only.
Teleconference.2008.01.16/Minutes
These minutes have been approved by the Working Group and are now protected from editing. (See IRC log of approval discussion.)
See also: IRC log
Contents
- Present
- Bijan Parsia, Ivan Herman, Alan Ruttenberg, Peter Patel-Schneider, Sandro Hawke, Ian Horrocks, Michael Smith, Boris Motik, Martin Dzbor, Carsten Lutz, Doug Lenat, Bernardo Cuenca Grau, Achille Fokoue, Jeremy Carroll, Zhe Wu, Michael Schneider
- Regrets
- Rinke Hoekstra, Uli Sattler, Deborah McGuinness, Jim Hendler, Ratnesh Sahay, Evan Wallace
- Chair
- Ian Horrocks
- Scribe
- Martin Dzbor
(Scribe changed to Martin Dzbor)
Ian Horrocks: we need to set this scribing business up on mailing list before the call...
Agenda amendments
PROPOSED: accept last week's minutes as per http://www.w3.org/2007/OWL/wiki/Teleconference.2008.01.09/Minutes
Ian Horrocks: do the minutes need more cleaning up?
Ian Horrocks: let's put it off until next week, ask Jim to clean it up
Ian Horrocks: let's treat the minutes as not accepted yet, we return to them at the next call
Ian Horrocks: Manchester minutes are also postponed for later? or not?
ACTION: Alan to add links to presentations to F2F minutes
PROPOSED: Accept F2F1 minutes with presentation links being added later
Alan Ruttenberg: links and presentations were communicated via mailing list
PROPOSED: Accept F2F1 minutes with presentation links being added later
RESOLVED: Accept F2F1 minutes with presentation links being added later
Pending review actions
Ian Horrocks: most are in a good shape
Ian Horrocks: Bijan has completed his action, Bijan comments?
Bijan Parsia: three actions concluded, scenario is under control...
Ian Horrocks: action for Jeremy, also completed?
Ian Horrocks: actions were completed adequately, so are there any problems with them? or should we close them?
? (guest): what we need to the with these actions pending review if we're happy with their conclusion?
Overdue Actions
Sandro Hawke: working on Action 43, in progress... no specific date yet
Sandro Hawke: will try to fix it by Jan 25th (?)
Ian Horrocks: Action 56 overdue is probably done -> Bijan says: has been rolled into n-ary proposal...
Bijan Parsia: action should be closed, more info to be circulated this week
Proposals to resolve issues
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
Ian Horrocks: peter refreshes what proposal is about
Peter Patel-Schneider: proposal is about moving from owl:DataRanges (as in current RDF mapping doc) to express user's datatypes to model them as instances of the RDF Schema class rdfs:Datatype; we checked rdf semantics, datarange is deprecated and rdfs:Datatypes makes it more consistent with the semantic web stack
Jeremy Carroll: we used data range before, we may want to think about the past uses before changing this...
Jeremy Carroll: only noticed now, owl:DataRange is also used for sets of plain literals... see:
Ian Horrocks: let's have a look at the next item, Issue 73, that seems to be clearer...
Issue 73: (Should owl:Thing be necessarily infinite?) as per http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0059.html
Ian Horrocks: this seems to be unanimous, people agree to reject
PROPOSED: close (as REJECTED) Issue 73 (Should owl:Thing be necessarily infinite?) as per http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0059.html
Jeremy Carroll: if nobody can support Jeremy's position, Jeremy notes his opposition, but sees no point in continuing debate
Bijan Parsia: is Jeremy's point an objection from HP? or what is its status...
Jeremy Carroll: it's not a formal objection
Ian Horrocks: close the issue, record Jeremy/HP voting against...
Jeremy Carroll: maybe HP should review the vote against, but perhaps at the next publication stage
Ian Horrocks: due to an opposing vote, make sure we record everybody's voice, incl. abstaining
RESOLVED: close (as REJECTED) Issue 73 (Should owl:Thing be necessarily infinite?) as per http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0059.html
Ian Horrocks: moving to Issue 74, (Use the xsd namespace for the facet names) as per http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0050.html
PROPOSED: close (as RESOLVED) Issue 74 (Use the xsd namespace for the facet names) as per http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0050.html
Ian Horrocks: discussed last time, everybody seems happy along the above lines
Ian Horrocks: quick vote
Michael Smith: asking about the scope... there might be some contentious items...
Ian Horrocks: discussed last time,
RESOLVED: close (as RESOLVED) Issue 74 (Use the xsd namespace for the facet names) as per http://lists.w3.org/Archives/Public/public-owl-wg/2008Jan/0050.html
Discussion: 1. Blank nodes and Skolems (Issue 3)
Ian Horrocks: some discussion on the mailing list... Bijan, Jeremy to sum up?
Bijan Parsia: responded to Jeremy's msg.,
Bijan Parsia: starting from the beginnning, in owl 1.0 we could represent tree-like patterns of anonymous existentials...
Bijan Parsia: serialized as blank nodes
Bijan Parsia: RDF extensionally qualifies all b-nodes, which in OWL would mean undecidability
Bijan Parsia: possibly have them as syntactic sugar, to handle RDF graphs as users expect; more RDF graphs would be DL legal and the behavior of the reasoner would be more familiar
Bijan Parsia: Carsten proposed a spec. 'total' role to represent true extensionals; algorithm is for SROIQ, and it is correct. SROIQ has the true universal role.
Bijan Parsia: objections, question?
Jeremy Carroll: discussed issue with colleagues, and it's clear that many implementation techniques use skolems... some may fail on entailing from bnodes... such failures not included in the conformance of the reasoner to the OWL semantics...
Jeremy Carroll: there were implementations in previous wg, that's why we defined certain checks only on consistency to get conformance... diff methods of implementation, can't mandate one...
Ian Horrocks: is this the case that if we define conformance w.r.t consistency check, not being able to tell diff
Jeremy Carroll: basically yes, diff parts of rdf graph may correspond to dl constructs
Jeremy Carroll: b-nodes treated diff in owl-full = its semantics becomes more difficult
Bijan Parsia: not understanding what is meant by impl. techniques, skolemization is a common one, but we're proposing something stronger, so that some entailments hold,
Bijan Parsia: if we reason about abox, we recognize the items as individuals, as people expect should happen
Bijan Parsia: owl-full seems a bit hypothetical problem?
Bijan Parsia: need to reconcile arguments that in some cases keep the old approach vs. cases that change semantics (of owl full), which could break reasoners
Bijan Parsia: we're trying to bridge the gap between dl and full, at the cost of some abstract parts of the theory being left out
Boris Motik: reiterating Jeremy, whether we treat skolems as individuals or exist. variables, doesn't matter... as long as we maintain consistency criteria? correct?
Boris Motik: we may actually strengthen this, in rdf we can have b-nodes on syntax level
Boris Motik: what about b-nodes in entailed ontologies?
Boris Motik: treat b-nodes as existentials?
Boris Motik: if it's on the right hand side, possibly... if it's in the graph part, it's something else
Ian Horrocks: interesting thing = worried about the case when everything is skolemized, incl b-nodes?
Jeremy Carroll: coming from owl-full implementation has some consequences... shouldn't we be worried that owl full would make non-entailments if this was adopted?
Ian Horrocks: when skolemizing individuals would cause problems... it's a thing what other people talk about?
Bijan Parsia: thought that in owl-dl there is a way of skolemizing syntax... we already use, everybody using rdf is already using those things
Bijan Parsia: in SPARQL, b-nodes can be in answers and can be treated as skolems...
Bijan Parsia: supporting sparql syntax over ontologies, it matters a lot how we treat it
Bijan Parsia: might not be variables, just funny renaming conditions
Bijan Parsia: this is an example of rdf syntax which is highly visible and has impact
Alan Ruttenberg: wants to clarify = concern is because of role implementation that generates entailments in owl-full... would be incorrect if skolemization is used?
Ian Horrocks: less complicated if considering skolemizatin as syntax?
Jeremy Carroll: answer boris... how jena treats b-nodes on the RHS
Jeremy Carroll: implemented to satisfy test cases = skolemized on the LHS, on RHS it's variables .......
Boris Motik: why conversion of RHS in sparql? is the diff betwen bijan's and jeremy's view in schema part?
Bijan Parsia: data counts... users may work on expectation that we take RHS and LHS, maintain mapping, entail, get them back
Bijan Parsia: no issue with schema part...
Jeremy Carroll: main concern is to change underlying semantics
Jeremy Carroll: not sure how changing semantics changes classificationa apps... why change semantics?
Ian Horrocks: classification may not make entailment visible... only uses satisfiability
Boris Motik: assume we're defining b-nodes in data as skolems... how to implement? not a matter of changing semantics doc, maybe a change for parsing doc?
Boris Motik: every b-node should be mapped onto an element of graph -> individual, etc.
Boris Motik: this may provide solution = on semantic level, b-nodes are existential vars, but there can be a switch to say that when parsing rdf, they should be treated as skolems... up to user/app to decide... could be?
Alan Ruttenberg: how this is visible... practical reasons, removes requirement and b-node as a tree, allows more flexibility,... intended meaning in most cases is skolemization...
Alan Ruttenberg: we don't see diff between skolems and existentials in owl 1.0... we need a negated property on individuals to see it??
Ian Horrocks: doesn't seem to be so diff... skolemizing on RHS should be visible even in owl 1.0
Bijan Parsia: agrees with alan that they are user-visible
Bijan Parsia: Q to Jeremy = what means "compelling argument" for/against something? e.g. many users wanting to use rdf graphs in owl reasoners, this is a powerful case, why not compelling
Bijan Parsia: similarly, sparql relationships, etc. = these are fairly important cases
Jeremy Carroll: to bijan first... what I haven't seen is how reworking semantics affects skolemization, is it necesary? (???)
Bijan Parsia: how to propose arbitrary b-nodes from rdf, what should they give/
Jeremy Carroll: no answer at this point...
Jeremy Carroll: to Boris.... finding that position of a value, at this moment...
Jeremy Carroll: there is some rationale behind it
Jeff Pan: maybe somebody can explain what this means for the users/end users... so that more people can join the debate?
Ian Horrocks: lot of explanation of this, maybe a bit technical...
Jeff Pan: if people write owl 1.1 axioms, what are the key diffs for THEM
Ian Horrocks: could Jeff, Jeremy prepare some examples? or Boris? or someone else?
Ian Horrocks: some examples would be useful to explain people what this is about
ACTION: Jeff to lead effort on formulating some examples on b-nodes issues and their impact on users
Ian Horrocks: good there was lot of effort on this.. and also some new ideas, suggestions
ACTION: Jeremy to respond to boris's parsing idea by e-mail
Issues
Ian Horrocks: no raised issues on agenda, no editorial either
Ian Horrocks: some to move on...
Ian Horrocks: ...syntax for allDisjoint... seemed simple, but complicated that if we have mapping for this, we should also have them for...
Peter Patel-Schneider: no need to go further in this discussion, syntax adapted to those few things that can be used a lot, not proposing extra syntax that do not require this
Peter Patel-Schneider: seems to be a reasonable resolution, unless people disagree
Ian Horrocks: maybe we should resolve it?
Jeremy Carroll: which special constructs do we mean?
Ian Horrocks: difIndividual, sameIndividual, etc.
Ian Horrocks: for allDisjoint there is aconstruct in semantics, there is nothing in mapping doc?
Peter Patel-Schneider: all disjoints always had structural.abstract syntax...
Peter Patel-Schneider: all we want is special for allDifferent
Ian Horrocks: difficulty with proposed solution is that if there is only structural syntax, no corresponding serialization, we may get to round-trippable problem
Ian Horrocks: what we have now is what we had in owl 1.0
Jeremy Carroll: disjoint Obj Props is new here...
Peter Patel-Schneider: started with a fact that something that needs to be expressed... where it got complicated is when we brough round-tripping... should we be accountable f
Boris Motik: what is peter's proposal? drop round-tripping, extend vocabulary? what is against extended vocabulary?
Peter Patel-Schneider: should we give up on round-tripping... to some extent, yes
Ian Horrocks: what's the problem of adding exactly that construct in mapping syntax? what is against it?
Ian Horrocks: people have diff opinions, there was a lot discussion on this
Alan Ruttenberg: isolate changes necessary to round-tripping...
Alan Ruttenberg: treat that separately and we can pay more attention to it
Boris Motik: is round-tripping important? pls. vote....
Ian Horrocks: seems to be worthwhile investigating round-tripping = action?
Ian Horrocks: initial input to who is affected, what impact it has, ....
ACTION: Boris to look at the round-tripping problem and collate initial material for/against it
Ian Horrocks: should we resolve Issue 2 and go for a new issue dedicated to round trippping?
Discussion on Issue 51 (quick)
Ian Horrocks: language name... some de-facto decisions
Ian Horrocks: called it owl 1.1.... any thoughts?
Peter Patel-Schneider: wanted to argue for 1.1.....
Alan Ruttenberg: we chose 1.1 to continue as a product line
Alan Ruttenberg: is this still an issue? esp. only Jim seemed to have objected...
Ian Horrocks: straw poll on what people think now?
Alan Ruttenberg: people asked whether this is still an issue.... people 'vote'
Bijan Parsia: can live with 1.1... maybe just call it OWL... maybe hard to define how much needs to go into ".1"
Jeremy Carroll: this is perhaps a policy question... should we use "owl" as a technology, or is this a "separate" technology "owl1.1"? we should see w3c position
Ian Horrocks: any other business?
AOB
Ian Horrocks: agenda for next week already available online
Ian Horrocks: agenda has quite some info in, so there is time to look at it and think about issues, discussion
Ian Horrocks: next week talking about punning, so pls. look at that
Alan Ruttenberg: suggests a check on issue times...
Ian Horrocks: concluded.....