See also: IRC log
- Peter Patel-Schneider, Boris Motik, Martin Dzbor, Rinke Hoekstra, Bijan Parsia, Uli Sattler, Diego Calvanese, Ian Horrocks, Michael Smith, Evan Wallace, Markus Krötzsch, Jeremy Carroll, Alan Ruttenberg, Jie Bao, Elisa Kendall, Sandro Hawke
- Achille Fokoue, Carsten Lutz, Ivan Herman, Michael Schneider
- Alan Ruttenberg
- Martin Dzbor
(Scribe changed to Martin Dzbor)
Alan Ruttenberg: we're starting
Alan Ruttenberg: F2F - discussed this with Ian and we're settling on 28-29 July at MIT, details pending
Alan Ruttenberg: want to include a topic on issues raised and how they are handled
Alan Ruttenberg: talking to IanH and things are going to change how issues are handled, raised,...
Alan Ruttenberg: there was an email about first batch of issues to review
PROPOSED: the above actions to be considered done
Bijan Parsia: My overdue actions got siderailed by the easyclasskey discussion so remain undone
Bijan Parsia: should be done in a few days
Jeremy Carroll: to review doc on OWL compatibility with RIF - prob. mid May - in line with their schedule
Alan Ruttenberg: are minutes acceptable?
PROPOSED: Accept Previous Minutes (from 16 April)
RESOLVED: Accept Previous Minutes (from 16 April)
Alan Ruttenberg: this section will be timed to 30 mins and we will return to issue later if time permits
Alan Ruttenberg: Open/Shut per Jeremy's note (http://lists.w3.org/Archives/Public/public-owl-wg/2008Apr/0198.html)
Alan Ruttenberg: this looks like minor change, not related to OWL
Alan Ruttenberg: considered editorial, already updated to reflect the issue
Alan Ruttenberg: any further issues?
Proposals to Resolve Issues
Issue 22 - Add sugar for rule (?x :hasSibling ?y) ^ :Male(?y) => (?x :hasBrother ?y)
Alan Ruttenberg: proposals to resolve issues
Alan Ruttenberg: we suggest to close it with no action
Markus Krötzsch: While rules are syntactic sugar, many applications of rules are prevented by current structural restrictions, even though the offending rules are not problematic in principle. We should be aware of that restriction. (see also this email) Alan Ruttenberg: does this problem affect specification? Markus Krötzsch: if you want to express... Jeremy Carroll: maybe we should postpone these issues to future WG-s? Uli Sattler: agrees to close this issue Uli Sattler: maybe consider this in the next extensions to OWL
Uli Sattler: some work can be done, but not really to change it all now
Alan Ruttenberg: nothing to do on this now...
PROPOSED: Consider Issue 22 formally closed as postponed
RESOLVED: Consider Issue 22 formally closed as postponed
Issue 57 - Errata in OWL 1.0 documents
Alan Ruttenberg: some errors spotted in OWL doc, since they are not worked on, we should close or postpone this
Alan Ruttenberg: suggests postponing
Jeremy Carroll: probably postpone is better
Evan Wallace: postpone is better
Bijan Parsia: None of the errors show up in OWL2 yes?
Alan Ruttenberg: strawpoll on postpone vs. close
Alan Ruttenberg: let's close it then
PROPOSED: Issue 57 Errata in OWL 1.0 documents closed as moot (not relevant)
RESOLVED: Issue 57 Errata in OWL 1.0 documents closed as moot (not relevant)
Issue 106 OWL 2 namespace per proposal
Alan Ruttenberg: what about namespaces? we keep old one, so OWL2 has the same namespace as the old OWL...
Boris Motik: when we mention namespaces, are we talking about RDF only or also about XML?
Alan Ruttenberg: only about RDF
PROPOSED: Resolve Issue 57 by saying that new OWL 2 vocabulary goes in old OWL 1 namespace
ACCEPTED: Resolve Issue 57 by saying that new OWL 2 vocabulary goes in old OWL 1 namespace
Issue 63 Defining an RDFS compatible semantics
Alan Ruttenberg: not sure if to accept this as a real issue
Alan Ruttenberg: should we close it as completed?
PROPOSED: Close Issue 63 as done.
ACCEPTED: Close Issue 63 as done.
Issue 119 Russel paradox in OWL Full due to self restrictions.
Alan Ruttenberg: this is an OWL Full comprehension issue
Jeremy Carroll: to members of previous OWLWG, whether there is new evidence to reconsider comprehension principle?
Peter Patel-Schneider: relationship between DL and Full needs rethink...
Peter Patel-Schneider: ...if comprehension principles go away
Alan Ruttenberg: what do we lose if we go the way Jeremy proposed?
Jeremy Carroll: deleting all comprehension principles allows us to rethink all relations, theoretically good and interesting, but practically that might not be valuable?
Peter Patel-Schneider: one can do a patch to keep things the same...
Peter Patel-Schneider: that might be adequate
Alan Ruttenberg: action on Peter or Ian?
Ian Horrocks: agrees with Peter, needs convincing that we should completely revise OWL Full semantics
Ian Horrocks: this might be closed to "out of scope" w.r.t. our charter
Bijan Parsia: wouldn't mind a new owl full, whether we can agree on the new version is a different issue
Alan Ruttenberg: someone should coordinate, sheperd the patching process
ACTION: IanH to sheperd/coordinate the patching process (per Issue 119)
Issue 97 Add GRDDL to OWL/XML Syntax
Alan Ruttenberg: anyone willing to take this issue?
Bijan Parsia: wants to say a thing on GRDDL
Jeremy Carroll: thinks we need the XSLT
Bijan Parsia: this seems to be almost editorial, depends on what level we are looking at this...
Bijan Parsia: there might be issue with going only for XSLT transformations
Alan Ruttenberg: this is a new information; why don't we ...
Alan Ruttenberg: ...discuss it later when there are others like Ivan and Sandro
Jeremy Carroll: we may get into maintenance problem; can clearly state that text is normative
Bijan Parsia: with OWL API we already have pretty reliable implementation, no point in preferring one implementation or that it should be part of OWLWG deliverables
Bijan Parsia: OWL API tracks the OWL2 spec
Alan Ruttenberg: my concern was satisfied with GRDDL; the main thing now is an easy, compatible translation
Bijan Parsia: there was supposed to be spec and then possibly different implementations?
Jeremy Carroll: GRDDL addresses how to do transformation from XSLT 1
Jeremy Carroll: that was recommended
Alan Ruttenberg: there is a trick, if you want to create a transformation into a language, create a trivial XSLT that replaces the output with the translation... can be done
ACTION: alanr to explore whether a simple transformation via XSLT would work in this case and to show trick for how to *generate* an XSLT to create a GRDDL transform
Issue 111 - User intent signalling
Alan Ruttenberg: would be good to allow users to signal, flag that ontology should be interpreted as OWL1, DL, etc.
Alan Ruttenberg: in the past MIME type was rejected
Alan Ruttenberg: Sandro put an initial email, any comments?
Boris Motik: idea was good, we need to include in the doc a switch to use given OWL semantics
Boris Motik: given RDF doc, we need to see under what OWL it is interpreted (DL, Full,...)
Boris Motik: maybe we don't need a switch per each profile
Jeremy Carroll: correction - MIME wasn't rejected in principle, just because there was no time
Alan Ruttenberg: from notes - this was not needed at the time
--> Bijan Parsia: one problem with MIME - tends not to work offline
Bijan Parsia: not robust enough
Bijan Parsia: for profiles - one reason for including is that people may want to signal that a mixed ontology is interpreted in appropriate way (say DL)
Bijan Parsia: not necessary to disambiguate semantics, but good to flag if users want to emphasize
Alan Ruttenberg: what happens when doing imports and there are different intents in different files
Sandro Hawke: mentioned in email, but no good answer
Sandro Hawke: not sure anybody knows how to implement certain combinations = may need to ban certain combinations?
Alan Ruttenberg: in solving this issue, we need to discuss and know what is the behavior
Jeremy Carroll: maybe we should have no semantics involved here, just graph (?)
Jeremy Carroll: to understand triples if we want to have just a graph, you can't use RDF semantics, as graph has no semantics
Bijan Parsia: Ok, in case of incompatible profiles, then a warning to the user should be signaled and the user offered a choice of which semantics to use
Alan Ruttenberg: kind of like intended semantics RDF entailment
Alan Ruttenberg: specific suggestions from bijan, sandro, jeremy? discuss on email and come back with revised proposal
Alan Ruttenberg: out of time... on this section and discussion
General discussion: Easy keys
Evan Wallace: Is the top property discussion postponed?
Uli Sattler: in the easy keys proposals we described a few things more explicit
Uli Sattler: semantics more explicit, explained why "easy", why this would cause problems
Bijan Parsia: there is one raised, open issue - depends on what we do about keys and b-nodes (?)... if variables can't go with easy keys
Alan Ruttenberg: can you collect situations we are trying to accomplish here?
Bijan Parsia: those will be included on easy keys page
Boris Motik: spoke to uli and got explanations, so no more reservations left
Alan Ruttenberg: we don't have Achille and Zhe - implementers and their position on this
Bijan Parsia: the intent of easy keys was to make it easy for implementation; unless they start messing with datatypes
Bijan Parsia: should work with anything that follows datalog rules
Alan Ruttenberg: bijan can you add a note about the case you worry about or explain it
Bijan Parsia: when you thing about dl-safe rules and datalog, you may not have some things (e.g. negation) in the right place, maybe boris, uli can tell more
Bijan Parsia: if you have finite datatypes you may have more work to do
Alan Ruttenberg: Jeremy is also among implementers - any issues?
Alan Ruttenberg: would be good to have affirmative responses from three implementers so that we can report on this
Alan Ruttenberg: action is needed on this
Alan Ruttenberg: ...would be good getting response from all implementers on the WG
ACTION: alan will get responses from Achille, Zhe, Jeremy as implementors - review of Easy Keys proposal
Alan Ruttenberg: other comments on easy keys or a straw poll?
Alan Ruttenberg: should we go ahead and have this feature added?
Alan Ruttenberg: strawpoll - are we ready to add this to spec
NONBINDING-PROPOSED: Are we ready to include EasyKeys to OWL specification?
Alan Ruttenberg: looks good, good work from bijan and uli to have proposal in such a good shape
Alan Ruttenberg: back to issue list
Alan Ruttenberg: more discussion on any previous issues
Issue Discussions (continued)
Issue 71 - Data ranges for literals with Language
Alan Ruttenberg: Jeremy not here, anybody has opinions on this - should we be handling XML literals as well?
Bijan Parsia: some variant on extending datatype language is fine, to do it somehow - may be useful
Bijan Parsia: XML literals are a bit dodgy - there is some form of inheritance... we should not support this, we don't support many other XML subtyping, because they are a hard problem
Bijan Parsia: go for a middle ground
Peter Patel-Schneider: suggests facet for this purpose
Peter Patel-Schneider: you don't have to impose syntax...
Alan Ruttenberg: will you write something for spec how this may look?
Peter Patel-Schneider: response sent to jeremy, may contain enough info
Alan Ruttenberg: jeremy was suggesting wildcarding? is that covered?
Alan Ruttenberg: e.g. EN* not necessarily, EN-GB EN-US
Bijan Parsia: if we want this from scratch, we can represent this using schema - string + values (facets can go into string part)
Bijan Parsia: this is not a fundamentally difficult thing, just needs to be in a neat way
Bijan Parsia: we should be able to search, cluster strings and in principle it's about having pattern facets
Alan Ruttenberg: what is the status of this right now?
Alan Ruttenberg: will it be easier to use patterns? bijan, will you spec ideas?
Alan Ruttenberg: other comments?
ACTION: Bijan to work out syntax for langed literals
Issue 112 Top property
Alan Ruttenberg: someone who understands "top" property to update us
Uli Sattler: we were unsure how to call it, but for some reason we needed "bottom" role, there was some discussion on universality
Uli Sattler: reasoners can get away with faking this
Alan Ruttenberg: bottom role - when debugging ontologies, one thing that came with unsatisfiable roles was that it could not have any values - could this be equivalent to a bottom role (e.g. maxCard 0)
Uli Sattler: might be a way to fake implementing bottom properties
Uli Sattler: maybe this needs syntactic sugar to name this
Alan Ruttenberg: would this add burden to reasoners?
Uli Sattler: not really, reasoners can handle this
Alan Ruttenberg: what the name should be?
Alan Ruttenberg: there are few possibilities, anybody?
Uli Sattler: some in the email posted recently...
Uli Sattler: should mirror owl:thing
Uli Sattler: "relatesTo" implies meaning which it doesn't really have
Alan Ruttenberg: the usual way to read properties is like a sentence - A topproperty B... sounds like relation between them
Uli Sattler: topProperty should related every individual with every other individual
Uli Sattler: "relatesTo" may be too ambiguous
Alan Ruttenberg: no problems with this
Alan Ruttenberg: strawpoll on adding top and bottom roles to OWL 2
NONBINDING-PROPOSED: Add "top" and "bottom" by some name, to OWL 2 specification?
Alan Ruttenberg: we have good sentiment that this is worth adding, let's discuss the actual names on email
Alan Ruttenberg: one last thing
Alan Ruttenberg: where we are going on whether there should be 1 or 2 OWL semantics documents...
Alan Ruttenberg: should there be one or separate
Alan Ruttenberg: adjourned