W3C * T&S * Semantic Web * WebOnt * prev: Jan NJ, Apr AMS

Third Meeting of the W3C Web Ontology Working Group

1-2 July 2002
Hosted by Stanford and Fujitsu in Stanford, CA

where/when/logistics * who * what: preparation, agenda, minutes/records

Where: Local Logistics Hotel info, etc.

Deborah McGuinness and Jonathan Dale and arranging local logistics.

see also:

Who: Participants

Per charter, participation is open to WG members and experts invited by the chair.

Registration closed 21 Jun.

The following WG members attended:

  1. Dan Connolly, W3C
  2. Jos De Roo, Agfa-Gevaert N. V.
  3. Michael Smith, Electronic Data System (EDS)
  4. Jonathan Dale, Fujitsu Limited
  5. Michael Sintek, German Research Center for Artificial Intelligence (DFKI) Gmbh
  6. Guus Schreiber, Ibrow
  7. James Hendler, Maryland Information and Network Dynamics Lab at the University of Maryland
  8. Marwan Sabbouh, MITRE Corporation
  9. Evan Wallace, National Institute of Standards and Technology (NIST)
  10. Ian Horrocks, Network Inference
  11. Larry Eshelman, Philips Electronic N.V.
  12. Deborah McGuinness, Stanford University
  13. Mike Dean
  14. Jeff Heflin
  15. Pat Hayes
  16. Raphael Volz, Forschungszentrum Informatik (FZI)

Regrets: Klein, Carroll, Olivry, Iannuzzelli, Patel-Schneider, Marchiori, Obrst, Buswell, Horan, Hellman, Miller, Gibbins, De Roure, Borden, Stein, Shimizu, Ned Smith, Said Tabet. Finin, Decker

Additional real-time info about who's registered is member-confidential.

What: Preparation, Goals

ftf meeting materials due 24Jun (per 13Jun telcon).

What: Agenda

From Guus's message of 19Jun; see also 25Jun from the host re social event Sunday evening.

SUNDAY JUNE 30

20:00 (or whenever) - Optional no host gathering at a local wine
tasting/cafe/restaurant in Palo Alto.
Niebaum-Coppola - http://www.paloaltoshop.com/pages/niebaum.html , 473
University Ave., Palo Alto, California. Phone: (650) 752-0350

MONDAY JULY 1

09.00 - 09.30    Setup, logistics, meeting objectives, agenda amendments

09.30 - 11.00    OWL Feature synopsis

Issue: what is in layer 1, what is in layer 2
Issue: What do we call layer 1
Issue: Resolution of pending naming issues
Resolve: to release this document as a WD

11.00 - 11.30    COFFEE BREAK

11.30 - 12.30    OWL reference description (part 1)

Issue: Raise any naming issues (last chance before release)
Issue: What requirements are still unmet - paragraph/appendix to address
Resolve: to release this document as a WD

12.30 - 14.00    LUNCH BREAK


14.00 - 15.00    OWL reference description (cont.)

15.00 - 15.30    "Extra-logical" features (part 1)

Discussion: numerous of our requirements require "extra-logical"
features - do we have a general mechanism for approaching these
(technical, not process), i.e. 
- import, 
- versioning, 
- ontology metadata,
- internationalization and display features, 
- extensibility mechanism

15.30 - 16.00    COFFEE BREAK

16.00 - 17.30    Extra-logical features (cont.)



19:00 - 22:00   Webont Social Dinner - Caffe Riace -
http://www.cafferiace.com/ . Address: 206 Sheridan Ave., Palo Alto,
California. directions. Phone: 650/328-0407.  No host.


TUESDAY JULY 2

09.00 - 09.45    Telecon
Zakim (tel:+1-617-6200), code 9326

09.45 - 11.00    OWL Formal Specification

Issue: Raise any naming issues (last chance before release)
Resolve: to release this document as a WD

11.00 - 11.30    COFFEE BREAK

11.30 - 12.30    GUIDE documents

Discussion: Our WG needs to produce documents to help users -
walkthru or the like.  Who, what, when?

See also proposal for target GUIDE results:
http://lists.w3.org/Archives/Public/www-webont-wg/2002Apr/0266.html
Plus discussion on  Apr25 telecon (agenda item 7):
http://lists.w3.org/Archives/Public/www-webont-wg/2002Apr/0339.html

12.30 - 14.00    LUNCH BREAK

14.00 - 15.00    TEST work

Discussion on test approach and plan. 

Jeremy's report on TEST session at A'dam ftf:
http://lists.w3.org/Archives/Public/www-webont-wg/2002Apr/0156.html
Ian's message on test cases
http://lists.w3.org/Archives/Public/www-webont-wg/2002Jun/0018.html

15.00 - 15.30    [Slot for unfinished discussions]

15.30 - 16.00    COFFEE BREAK

16.00 - 17.30    MISC

- Planning: until Bristol ftf
- Resolve: Release updated Requirements Document

What: Minutes/Records

MONDAY JULY 1

1. OWL 1.0 Feature synopsis

1.1 Naming issues

[Resolutions affect all three documents]

Raphael Volz has sent a list of differences between the three writeups, this is distributed to participants.

Below is the set of actions and resolutions wrt. these naming issues.

RESOLVED: Norms for capitalization

ACTION DebM: in 2.2, 2.2, 3.1, 3.2 use a distinctive font for terms in the language and conventional capitalization

ACTION Deb: clarify that owl: is the default, note when names come from RDFS

RESOLVED: owl:Ontology is an agreed term, as used in OWL 1.0 (it is Ok that feature synopsis doesn't have it, since feature synopsis needn't be exhaustive, esp. w.r.t. non-logical)

RESOLVED: owl:versionInfo likewise

PROPOSED: owl:imports
RESOLVED (Connolly, Hayes, Horrocks abstaining. Dean opposed)

RESOLVED: (at earlier telecon): owl:allValuesFrom, owl:someValuesFrom (note plural)

ACTION Horrocks to ensure a section is added to the asbtract syntax doc with a mapping from the OWL reference doc names to the (additional) names used in abstract syntax (e.g. EquivalentClasses).

PROPOSED: to normalize all three docs to owl:cardinality, owl:minCardinality, owl:maxCardinality
RESOLVED

PROPOSED: in OWL Lite to go with owl:cardinality/min/max, restricted to "0" or "1",
RESOLVED (dissenting: Wallace; abstaining Hendler, McGuinness, de Roo, Connolly)
ACTION DebM to update feature synopsis, restricted cardinality section
ACTION MikeD: note owl-lite restrictions on cardinality in reference doc

RESOLVED: owl:intersectionOf.
ACTION Ian update formal spec.

ACTION DebM: note ObjectProperty, DatatypeProperty in the note about "datatypes TBD"
ACTION Ian: get IndividualProperty changed in formal spec doc

RESOLVED until we resolve issue 3.4 we use owl:UniqueProperty owl:UnambiguousProperty
ACTION DebM: update feature synopsis functional->Unique (noting 3.4) and isTheOnlyValueFor to owl:UnambiguousProperty (note the spelling of owl:UnambiguousProperty.)

ACTION MikeD: "sameIndividual in text" fix typo.

RESOLVED to thank Raphael for the detailed work.

1.2 Other (non-naming) issues

PROPOSED from Mike Dean - we allow sameClassAs only between named classes in the OWL lite case
RESOLVED

ACTION Ian: investigate the implications of striking "Class descriptions..." text from the document.

"Class descriptions can either be partial, indicating that the elements of the class satisfy at least the stated description and perhaps others; or the class description can be complete, indicating that the elements of the class are precisely characterized by the

PROPOSED: "OWL Lite" for the subset, "OWL" for the full language
RESOLVED, no objections, connolly abstaining

STRAW POLL: add a Disjoint feature to owl-lite one in favor, three against
RESOLVED: to leave disjoint out of owl-lite

ACTION add "Thing" to owl-lite description. add "Nothing" to owl (full) description

ACTION Deb to rewrite intro following Ian's comments (motivation should include full OWL)

PROPOSED: release feature synposis document as a W3C WD, with editorial changes. (contingent on availability of reference document) [For release schedule, see telecon on day 2]
RESOLVED

ASIDE: Recap of doc revision procedure:

2. OWL 1.0 Reference description

ACTION Mike Dean to include explanation of class elements and expressions

Discussion on overlapping/distinctive OWL/RDFS language features:: owl/rdf:Property owl/rdfs:subClassOf owl:rdfs:subPropertyOf owl/rdfs:domain owl/rdfs:range

STRAW POLL
rdfs:subclassof only: prefer 5 can live with 9
rdfs:class + owl:subclassof: prefer N-3; can live with: N-3
RESOLVED: For the moment we will go with the owl:subclassof format, as this was the DAML+OIL way

ACTION DanC to raise an issue wrt rdfs:subclassof and owl:subclassoff

ACTION editor to propose changed title, if deemed necessary

PROPOSED: to use for the namespace name: http://www.w3.org/2002/07/owl#
RESOLVED

ACTION Connoly to get the above OK'd by the W3C webmaster/director;
ACTION Dean to update the reference document

PROPOSED: to release OWL Reference document, taking inrto account the various editorial changes [For release schedule, see telecon on day 2]
RESOLVED

3. Extra-logical features

[no decisions, just initial discussion]

Raphael: should we define our "owl" version of Dublin Core features?
DanC: strongly agianst, tools like Adobe may have dificulties interpreting owl:title
Deb: confirms that her current wording about DC is OK no normative lagnauge features for ontology metadata

Raphael: use CVS format for versioning information Jeff: ability to state backward compatibility

Tagging of statements- must raise issue, remove from requirements, or decide it is already handled in some way

Internationalization: go with RDF although not big fan of it

Other issues that are solved: character model

Discussion about extensibility mechanism

SUMMARY: time needs to be spent on versioning and import language features Req: attachment of info to statements It will have to be clear tomorrow whether there is an issue here

TUESDAY JULY 2

4. Telecon

4.1 Roll call

Borden, ter Horst, Patel-Schneider, Stein, Welty plus ftf participants

4.2 Document release schedule

Discussion about time available for document review Sufficient consensus for extra week

PROPOSED: July 11: new versions; comments by July 15; decision by July 18 if we have consensus 18July, we can publish; else we take another week to decide
RESOLVED

4.3 AOB

Hendler summarizes discussion on extra-logical features
Patel-Schneider wants some current issues to be opened
DanC: requests that 5.10 be opened
PPS: Would like 5.3 and 5.10 referenced in language description
DanC: maybe formal spec should be where these issues are referenced

5. OWL 1.0 Abstract syntax (was: OWL 1.0 Formal Specification)

Two proposals for a model theory:

SEM: semantics for the abstract syntax Peter F. Patel-Schneider Tue, Jun 11 2000 http://lists.w3.org/Archives/Public/www-webont-wg/2002Jun/0082.html16:55:06

Re: layering (5.3,5.10): a first-order same-syntax model theory Dan Connolly (Fri, Jun 28 2002) http://lists.w3.org/Archives/Public/www-webont-wg/2002Jun/0209.html17:04:12

PROPOSED: add mapping to abstract syntax but no model theory release on same schedule as other documents Note: "no model theory" doesn't preclude adding one in this meeting
STRAW POLL:
in favor: 11
opposed: 1
opposed (DanC) can live with
RESOLVED: Abstain Danc, Jos, Jeff, ??

PROPOSED: define OWL by addition from OWL lite
STRAW POLL: In favor: 11, opposed: 3
RESOLVED: with 7 abstains "Formal Spec" is misleading, at least until it has a model theory
RESOLVED: ??

STRAW POLL: Put Peter's model theory through the standard editorial process
In favor: 8
Opposed: 5
Abstain: 3

Model theory break-out

At this point a break-out group was formed to discuss the model theory issues in more detail: Connolly, Hayes, Heflin, Horrocks, de Roo. A post-hoc summary by Hayes:

We argued a lot and discovered some previously unknown philosophical differences about the nature of meaning. The final positions were that the chief problem with Dan's semantics was that getting it right and discovering its properties was basically a research effort which we didn't have the charter or the time to undertake, in contrast to the traditional semantics techniques used in Peter's model theory. (Which is more an endorsement of Peter's style, as it were, than of the particular details, which we didn't spend much (any?) time on). On the other hand, Dan felt strongly that the OWL MT should be stated in terms of the actual OWL syntax (rather than any kind of 'abstract' syntax) and that it should be related as closely as possible to the RDF MT. To some extent, both sides seemed to be converging on the idea that it should be possible for everyone to be happier than they had been until now, if we agreed to put aside purely stylistic differences.

What I said back at the main meeting, and which Jim attempted to transcribe, was something like the following. This was in response to a question, whether we had made progress on layering. I said that

  1. there was light at the end of the tunnel, which was meant to convey the uneasy sense of guarded optimism that the breakout group seemed to have arrived at;
  2. it seemed to me that we could put together a reasonable overall story about layering OWL on RDF which would go as follows: every piece of OWL maps into (or simply is) a set of RDF triples, some of which may have to be 'dark', but that this can be kept to a minimum; however, not all sets of RDf triples are 'legal' OWL, so an OWL parser might have more work to do than an RDF parser. In particular, it may be that OWL will impose tighter semantic constraints on the non-dark triples (for example, OWL may forbid self-membership of classes). I also said these two kinds of layering imperfection are inevitable in some form or another, and we can live with them.
  3. I volunteered, as a step towards getting this clear, to try to take the current syntax (which defines a mapping into RDF triples) and Peter's model theory and try to use them together to define a model theory in "the Connolly style", by which I meant, stated in terms of constraints on the interpretation of sets of RDF triples rather than directly on the abstract syntax, in a way that would conform to the sense of optimism mentioned above. I also, rashly, promised to do this in a 3-week period which contained the 4th of July.

A few specific entialments were discussed (photos of whiteboard); e.g.

?x rdf:type FileTransAgent
FileTransAgent guardedBy ?G3
==> ?
?x rdf:type _:PolicyClass.
_:PolicyClass guardedBy ?G3.

and

?x rdf:type owl:Class.
?x my:prop "z".
==> ?
?x rdf:type _:something.
_:something my:prop "z".

ACTION Pat will attempt to take abstract syntax, and Peter's MT and the mapping into RDF and will write a model theory in the Connolly style (i.e. as an extension to RDF MT) and see if he can identify the exact issues. Target deadline: three weeks (July 25).

SCHEDILING NOTE: the model theory should be available for release after the Bristol ftf. This means a draft should be available by Sep 1.

6. GUIDE documents

Useful input:
http://www.xml.com/pub/a/2002/01/30/daml1.html -- by Roxane Ouellet, Uche Ogbuji
http://www.ksl.stanford.edu/people/dlm/papers/living-with-classic-abstract.html

6.1 Walkthru

Dtatatypes should be covered in the walkthru.

DAML+OIL document is basis. Should be updated with realistic examples.

ACTION Mike Smith to edit walkthru, with help of Lynn Stein, Ora Lassila, Deb McGuinness. Guus Schreiber to help with examples.

6.2 How-to-do-it document

Possibility of doing the "how-to-do-it" as a FAQ

Suggestion Evan Wallace: collect examples of language use on web page

ACTION Guus will generate a structure in which the examples should appear by July 11. This will also include one example.

ACTION Evan Wallace and Larry Eshelman: contribute examples

6.3 Presentation syntaxes

ACTION Evan will writeup a description of a recent OMG meeting that concerned UML and OWL, and the process he is running at OMG, and will post that to the WG.

7. Requirements document

Purpose:

R1 Ontologies as distinct objects

Raphael expressed concern with using rdf:about="" changes if you move the ontology2

ACTION (Raphael): send Guus paragraph suggesting preferred usage for owl:Ontology

R2 unambiguous term referencing with URIs

No known problems

R3 explicit ontology extension

Probably depends upon import issue. Currently no strong notion of ontology extension.

R4 commitment to ontologies

??

R5 ontology metadata

Currently can put anything in ontology headers.

ACTION (Mike Dean): use DC attributes in owl.owl

R6 versioning

Open issue

R7 class definition primitives

Addressed

R8 property definition primitives

Addressed

R9 data types

Open issue

R10 class and property equivalance

Addressed. Open issue to mix classes and properties

R11 individual equivalance

owl:sameIndividualAs

R12 local unique names

Motivates owl:differentIndividualFrom

Role of UnambiguousProperty?

ACTION (Deb) write up an issue with respect to the unique names assumption requirement

R13 attaching information to statements

Currently only mechanism is RDF reification,
Mike showed example of using RDF statementIDs to show that Deb's hair was red on Tuesday.
Not addressed by OWL model theory [mdean]
Jim suggests using owl:tag as an uninterpreted standard property, See issue 4.4 extra-logical feature set

R14 classes as instances

ACTION (Raphael) raise superissue to subsume equivalentTo
RESOLVED: close issue 4.6 equivalentTo to be subsumed by the new issue Raphael will raise

R15 complex data types

Pending issues

R16 cardinality constraints

Addressed

R 17-20

Satisfied with RDF solutions

RESOLVED: Release the new draft of the requirements document as is

8. TEST - WG Schedule

WG is still working in accordance to the revised schedule.

To move directly from Last Call to ProposedRec an abundance of implementation needs to be available.

ACTION Jim will work with Mike Dean to see if DAML Validator can become OWL validator

ACTION (all): send to WebOnt mailing list a short description of the tools you have available. or that you will use to help tools.

ACTION (Raphael): will make a large ontology available to test readers.

9. AOB

RESOLVED: Thanks to Jonathan and Deb for local arrangements - special thanks to Fujitsu for the wine and lunches - kudos to Deb for great choice of restaurant.

ADJOUNED at 17:29 Pacific time to leave time for a group photo.

IRC logs:

also: logs at ILRT.


Dan Connolly, for Guus S. and Jim H., chairs
$Revision: 1.34 $ of $Date: 2002/08/22 13:59:26 $ by $Author: connolly $