W3C Web Ontology Working Group IRC Chat Logs for 2002-01-14

09:33:03 Users on #webont: @logger

14:11:32 * em wonders if there is a webont channel for f2f?

14:11:34 <DanC> [Lucent host]: Sheila M and I are running a workshop...

14:11:52 * DanC is coming from the ftf; dunno whether this channel will be "official" in any way

14:12:04 * em waves to DanC

14:12:59 <DanC> agenda: http://www.w3.org/2001/sw/WebOnt/ftf1.html#agenda

14:13:05 <DanC> dinner tonight is at 6pm

14:14:25 <DanC> previously mentioned workshop: e-Services and the Semantic Web http://www-db.research.bell-labs.com/essw2002/

14:17:32 <DanC> ==== JimH convenes

14:17:41 <DanC> JimH models WOWG t-shirts

14:18:21 <DanC> === Roll Call

14:19:25 <DanC> Finin accepts scribe duties for the first session.

14:19:35 <DanC> -- Finin, UMD

14:19:50 <sandro> PFPS was making noises about making some of the session a telecon and inviting some of us not in the WG....?

14:19:52 <DanC> -- Connolly, W3C

14:20:02 <DanC> -- PeterPS, Bell Labs/Lucent

14:20:06 <DanC> -- IanH, U Mann

14:20:15 <DanC> -- Shimizu (sp?), INTAP

14:20:29 <DanC> -- Carroll, HP

14:20:32 <DanC> -- Smith, EDS

14:20:39 <DanC> -- Fensel

14:20:41 <DanC> -- Decker

14:20:47 <DanC> -- Volz

14:20:53 <DanC> -- McGuinnes

14:20:58 <DanC> -- Hefflin

14:21:04 <DanC> -- Hellman, Unicorn

14:21:09 <DanC> -- Obrst, Mitre

14:21:14 <DanC> -- Van Harmelen (sp?)

14:21:15 <DanC> -- Dean

14:21:17 <DanC> -- Gibbins

14:21:24 <DanC> -- Dale, Fujitsu

14:21:30 <DanC> -- Jos De Roo

14:21:36 <DanC> -- Olivry, EDF

14:21:44 <DanC> -- Sabou (sp?), Mitre

14:21:50 <DanC> [...]

14:21:51 <DanC> -- Miller

14:21:54 <DanC> -- Brickley

14:21:57 <DanC> -- Schriber

14:22:08 <DanC> [..] = Barnette

14:22:24 <DanC> JimH: one or two are still expected.

14:22:40 <DanC> here = irc.openprojects.net

14:23:03 <DanC> PeterPS: NOTE WELL: I'm not sure the logistics for getting in the room will be exactly the same tomorrow.

14:23:05 <mdean> mdean is now known as mdean_

14:25:08 <DanC> JimH: I've been asked "what are we expected to produce"?

14:25:18 <DanC> JimH: [cf charter, Director's Decision]

14:25:54 <DanC> DanC has changed the topic to: WebOnt ftf, NJ http://www.w3.org/2001/sw/WebOnt/ftf1.html

14:26:24 <DanC> JimH: we were not chartered to build, from scratch, a new language; we're chartered to build on existing stuff.

14:26:48 * DanC notes that the chair did designate this channel for minutes

14:28:21 <DanC> JimH: we're not doing content ontologies; contrast with SUO

14:28:53 <DanC> JimH: we're trying to find the suite spot between expressiveness and utility.

14:30:01 * DanC welcomes tim, our scribe for this session

14:30:27 <DanC> JimH: [1.2.2 Formal Semantics ...]

14:31:27 <tim> jimh: need to document how owl differs from other languages

14:31:50 <tim> jimh: this documentation needs to describe and be accessible to different user communities

14:32:21 <tim> question: who is the target reader of our standards documents?

14:32:41 <tim> jimh: developers plus users

14:33:03 <tim> jimh: look at the daml documents as examples

14:34:12 <tim> jimh: danc gets up

14:34:38 <tim> danc: we're here to get a technology deployed

14:34:51 <tim> danc: whatever is required to do this is what we should do

14:35:28 <tim> danc: experience in the xml schema group was that they needed two spec documents

14:36:13 <tim> danc: we'll go until 10am

14:36:57 <tim> danc: looking at http://www.w3c.org/2001/sw/WebOnt/

14:37:27 <tim> danc: look at background references

14:38:21 <tim> danc: look at http://www.w3.org/Guide/

14:38:55 <tim> danc: looking at http://www.w3.org/Consortium/Process-20010719/

14:39:22 <tim> danc: is describing W3C

14:40:53 <tim> jimh: we might need to redo the IP form for member orgs

14:41:39 <tim> danc: we're are chartered to produce a technology that can be implemented on a royalty free basis

14:42:15 <tim> danc: the format of w3c tech reports is fixed. we have to live with it. checked by machine.

14:44:48 <tim> danc: describing the "last call" part of the process

14:45:25 <tim> danc: any other working group or even a member of the public can raise issues to say whether or not a working group is "done".

14:45:58 <tim> danc: looking at http://www.w3.org/Consortium/Process-20010719/tr.html#last-call

14:47:27 <tim> pps: describing the networking environment for the meeting

14:48:51 <tim> danc: when we publish working drafts and in what form are up to us

14:49:18 <tim> danc: we are required to publish a working draft every three months

14:49:53 <tim> jimh: use case docs will come out as a working draft

14:50:11 <tim> danc: working drafts elicit feedback

14:52:59 <tim> danc: simplicity is a big survival characteristic for these kinds of standards

14:53:23 <tim> danc: a wg needs to stick around for ~6 months after the recommendation

14:55:54 <tim> danc: back to reviewing parts of http://www.w3.org/2001/sw/WebOnt/

14:56:53 <tim> jimh: xml-schema is not on this list, is this intentional?

14:57:48 <tim> comment: we need to be able to say how what we are doing is different from xml schema

14:58:02 <tim> danc sits down

14:58:17 <tim> pps: one more item before break -- the name.

14:58:33 <tim> jimh: we had a resolution on last telecon

14:58:38 <tim> pps: owl is fine

14:58:55 <tim> jimh: we resolved in the last telecon that we will call the language OWL.

14:59:24 <tim> danc: oit's as official as it will get before submitting the recommendation

14:59:56 <tim> jimh: doesn't change the name of the working group, webont

15:00:51 <DanC> s/submitting the recommendation/publishing a working draft/

15:03:25 <tim> ___

15:03:25 <tim> <*,*>

15:03:25 <tim> [`-']

15:03:25 <tim> -"-"-

15:03:41 <tim> danc: reviws schedule for rest of day

15:04:33 <tim> jimh: reviewing tomorrow's schedule

15:05:18 * DanC wonders if anybody's tring to call in

15:05:21 <tim> a call just came in on the conference call

15:05:34 <tim> we answerewd but no one was there.

15:05:42 <tim> perhaps it was a telemarkerer

15:05:54 <tim> jimh: we are now on break until 10:30

15:06:21 <tim> and so it goes

15:38:18 <DanC> ======= next session

15:38:29 <DanC> Frank vH accepts scribe duties for the session

15:38:50 <jah-mac> jah-mac is now known as jimhWowG

15:41:22 <DanC> did Ian accept an action to archive his presentation materials, JimH?

15:42:01 * DanC suggests everybody say a prayer to the demo-gods...

15:42:27 <tim> ?

15:42:54 * las is happy to say it works!

15:43:53 <DanC> logger, pointer?

15:43:53 <DanC> See http://ilrt.org/discovery/chatlogs/webont/2002-01-14#T15-43-53

15:45:14 <DanC> slide 4.

15:45:35 <DanC> Q: 1st bullet: there's lots of XML Schemas and b-to-b stuff going on...

15:45:56 <DanC> so when was that claim made?

15:46:41 <DanC> A:most stuff is still only human-readable.

15:46:56 <DanC> A:XML Schema approach seems to require pre-existing knowledge of what documents mean.

15:48:26 <DanC> hmm... does DAML+OIL have ABox/TBox?

15:48:45 <DanC> hmm... does it really have expressions? it's all just built with 2-place predicates, no?

16:04:25 <JosD> do we want to say: for this document we have the UNA ???

16:05:11 <DanC> JimH, are you keeping an eye here?

16:05:25 <DanC> I wonder if we could have a session with the RDF validator, which makes pictures from RDF syntax

16:05:47 <DanC> MikeDean, could I borrrow/steal some of your time tomorrow for playing with the RDF validator?

16:07:13 <em> DanC, do you have IsaViz loaded on your laptop? This may make more sense as you can interact with the images produced

16:07:21 <DanC> isaViz: nope

16:09:01 <DeborahMcGuinness-SWWS> slow response to josD - yes

16:17:15 <DeborahMcGuinness-SWWS> also users may want to apply the closed world assumption at some point on a knowledge base (similar to people wanting to make the unique names assumption for some knowledge base)

16:18:49 <DanC> re syntactic sugar: how about noting a minimal basis... for implementors, the terms that are sugar are much easier to implement.

16:21:03 * DanC asks Frank to be sure he gets this Q/A stuff

16:21:21 <DanC> Q: what's up with this xxxQ stuff? when do we use it?

16:22:20 * DanC realizes Frank isn't using this channel for his notes.

16:28:19 <DanC> Lynn: hasClassQ is a mechanism for making an n-ary predicate in RDF, yes?

16:28:22 <DanC> PeterPS: yes

16:29:27 <DanC> JimH, I think your point did get lost in the technical discussion; I think your point was: if we do something wierd (i.e. making a 5-ary relation in a syntax that's designed for 2-ary relations) we'll have to explain it specially.

16:29:51 <las> Specifically, hasClassQ AND minCardinalityQ (or max, etc.) together define a 5-tuple: subject, property=hasClass(Q), objectOfHasClass, property2=cardinality(Q), objectOfCardinality

16:30:45 <las> I agree that JimH's point was lost and apologize for distracting, but thought it was important (a) to make sure I understood and (b) if possible, to increase the number of other people in the room who thought they understood, to (or even thought correctly they understood :o) )

16:37:41 <DanC> yes, there is another important point there: RDF is heavily-biased to 2-ary relations. Doing n-ary relations hurts.

16:38:37 * sandro would argue that it only hurts once, and you quickly get used to the pain with a little syntactic sugar.

16:39:00 <sandro> As with LISP, which also only does binary relations, at the lowest level.

16:39:15 <DanC> hmm... I think the pain is pretty long-lived. It seems economical to push the pain down into the RDF layer.

16:40:00 * sandro also wouldn't mind chucking 3-tuples for an n-tuple model, if that's really an option,.

16:40:30 <sandro> but there may be drawbacks I haven't figured out, either. It's a big decision.

16:40:33 <DanC> yup

16:45:17 <jah-mac> jah-mac is now known as JimHWoWG

16:47:18 <JimHWoWG> JimHWoWG is now known as Jah-wowg

16:48:10 <DanC> hmm... is this Oiled thing OpenSource?

16:48:31 * DanC surfs around... http://img.cs.man.ac.uk/oil/

16:49:35 <DanC> hmm... http://img.cs.man.ac.uk/oil/license.html looks OpenSource-happy to me

16:50:21 <DanC> oops, no, it's not: "Permission is not granted to disassemble, decompose,

16:50:22 <DanC> reverse engineer, or alter this file or any other files in the

16:50:22 <DanC> package. "

16:52:54 <DanC> re decidability: I wonder if I should bring up that this isn't a requirement here.

16:53:08 <DeborahMcGuinness-SWWS> sometime that is worth discussing

16:53:21 <DeborahMcGuinness-SWWS> but the discussion will not be short

16:53:33 <DanC> s/isn't a requirement/isn't yet a requirement/

16:58:41 <DanC> hmm... having a top class seems like it would bring up Russel's paradox.

16:58:54 <DanC> i.e. the class of all classes that are not members of themselves.

17:00:15 <JosD> i.e. see http://www.agfa.com/w3c/euler/russell.axiom.n3 no?

17:01:00 <DanC> this abox/tbox stuff isn't in my background. Is it familiar to anybody else?

17:01:40 <las> Dan, it is standard (i.e., obscure) KR, but you are absolutely right that Ian should not be taking it for granted.

17:03:20 <las> Google reveals http://www.hpl.hp.com/semweb/download/DescriptionLogicsIntro.pdf (see esp. p. 10) when queried on abox tbox. Seems appropriate enough....

17:04:13 <las> (abox = assertions, i.e., ground facts; tbox = terminology, e.g., class relations

17:04:14 <las> )

17:05:40 <Jah-wowg> the hard part is remember which is which if you're not a DL person - I always forget if it is A for Axiom or T for Theorem or...

17:11:29 <DeborahMcGuinness-SWWS> we should add a pointer to this

17:11:49 <DeborahMcGuinness-SWWS> actually the distinction started to disappear in DLs after constructors like oneof and filler ended up on concept languages

17:11:55 <DeborahMcGuinness-SWWS> t box used to just be the schema

17:12:03 <DeborahMcGuinness-SWWS> and abox used to just be the instances

17:12:18 <DeborahMcGuinness-SWWS> once you could put things like clinton in the tbox with constructors like one-of

17:12:29 <DeborahMcGuinness-SWWS> most people in the dl community started dropping the distinction

18:00:18 <DanC> ====== lunch ends

18:02:07 * nmg scribes

18:02:54 <nmg> ====== Use Case Discussion

18:02:55 <mdean> mdean is now known as mdean_

18:03:19 <nmg> ====== Collection Management (Schreiber)

18:03:28 <TimFinin> ?

18:04:32 <nmg> danc: use case document motivates requirements

18:05:50 <nmg> jimh: requirements doc in three parts: 1 compelling use cases (5ish)

18:05:57 <nmg> 2 requirements arising from use cases

18:06:22 <nmg> 3 appendix containing additional use cases not contributing to 2

18:06:55 <nmg> jimh: requirements section is a 'living document'

18:08:51 <nmg> jimh: requirements arising from existing use cases not completely consistent

18:09:49 <nmg> jimh: purpose of this meeting is to reduce these to a consistent set

18:11:25 <DanC> DanC: what we're looking for in use cases is: can we use this as a finish line? i.e. does pretty much everybody in the group agree that "if we can't do that, we're not done."?

18:11:45 <DanC> JimH, has gus accepted an action to archive his presentation materials?

18:15:10 <nmg> guus: requirement arising from a common modelling issue: classes as instances of classes

18:15:59 <DanC> Velente et al... anybody else read it? pointer?

18:16:57 <DanC> [repeat, since maybe Deb can help]

18:16:58 <DanC> Velente et al... anybody else read it? pointer?

18:17:44 <nmg> ianh: how useful is this metaclass approach?

18:18:43 <DanC> re Valente, googling around produced http://citeseer.nj.nec.com/253542.html ; am I close?

18:19:00 <nmg> guus: this arises in both stephen buswell's aero example and jjc's arkive example.

18:19:33 <nmg> las: many systems, including frame and oo systems in which metaclasses are used in this way

18:19:53 <nmg> ianh: semantic are confused

18:20:31 <DeborahMcGuinness-SWWS> this could be the valente ref: (no hot link sorry - found on google - "Building and (Re)Using an Ontology of Air Campaign Planning&quot;. A. Valente, T. Russ, R. MacGregor, and W. Swartout. IEEE Intelligent Systems 14(1), 1999.

18:20:43 <nmg> las: existing literature and user community make use of this - may be 'wrong', but should be noted

18:20:50 <nmg> action: las to explain further

18:21:31 <DanC> if you would, please put ACTION in all caps. easier to search for

18:21:46 <nmg> guus: daml+oil and rdf allow it, but semantics are problematic

18:22:03 <nmg> guus: req #2: constraints

18:22:08 <DanC> nmg, there's no need to re-capitulate the presentation materials here.

18:22:14 <nmg> okay, noted

18:27:10 <nmg> danc: point of order - f2f useful for conducting polls

18:27:58 <nmg> danc: on both use cases and requirements

18:28:24 <nmg> danc: subgroup membership does not exclusively represent members' interests

18:29:27 <nmg> jimh: not for now - breakout sessions tomorrow

18:31:29 <nmg> jimh: as each requirement is presented, take straw poll of members, record roughly (strongly agree, etc)

18:32:45 <nmg> POLL: classes as instances of other classes (strong agree)

18:35:16 <nmg> POLL: definitional constraints (mostly unclear)

18:35:23 <DanC> no in favor.

18:35:56 <DanC> in particular: the x.length < y.length isn't (clearly) doable in DAML+OIL

18:39:47 <DanC> default: several in favor, several against...

18:40:04 <TimFinin> what was the default for the group?

18:40:56 <DanC> repoll

18:41:06 <DanC> default requirement yes: several

18:41:14 <DanC> not a req: many

18:41:39 <nmg> POLL: default knowledge (mostly in favour, some against)

18:41:59 <nmg> CORRECTION: default knowledge (mostly against, some in favour)

18:44:01 <nmg> las: individual decisions rely on other issues - votes often qualified

18:45:36 <DanC> hmm... a lot of his requirements look like rules to me.

18:47:33 <DanC> "his" meaning the ones he's presenting

18:48:31 <JosD> DanC, how so for part/whole

18:50:19 <DanC> well... parts of airplanes inherit from the whole, but parts of dressers don't. i.e. you need rules to say which is which. or: you can express the difference with rules.

18:50:27 <nmg> POLL: part/whole relations (mostly opposed, few in favour)

18:52:35 <JosD> DanC, yes that expressing of the difference is indeed an interesting idea!

18:53:42 <nmg> POLL: property typing (mostly in favour, couple against, few don't care)

18:54:41 <nmg> jimh: general action on all those presenting to provide electronic copy of presentation materials for archival

18:54:49 <nmg> ACTION: guus to provide slides

18:56:00 <nmg> ====== Content Interoperability (Leo Obrst)

18:56:18 <nmg> presentation materials are those mailed to www-webont-wg

18:56:59 <nmg>http://lists.w3.org/Archives/Public/www-webont-wg/2002Jan/0066.html

18:57:28 <DanC> more abstract: bummer. I like use cases to be concrete

18:59:37 <nmg> leo: requirements generalised from 22 use cases

19:02:51 <nmg> jimh: should not vote on each requirement - too many. leo should select key ones.

19:03:27 <nmg> POLL: inter-ontology references 3.1.1 (mostly in favour)

19:05:37 <nmg> POLL: ontology mapping rules, features 3.1.3 (jimh rules poll out of charter)

19:06:44 <DanC> hmm... that was the first time the chair curtailed discussion based on the charter. perhaps that bears explanation

19:06:46 <nmg> POLL: ontology composition language 3.1.4 (mostly in favour)

19:10:52 <nmg> POLL: inter-ontology sysnonyms/aliases 3.1.8 (mostly in favour, no against, few don't care/know)

19:13:22 <nmg> POLL: ontology approximation 3.1.11 (mostly against, few don't care, couple in favour)

19:17:26 <nmg> POLL: inter-ontology validation 3.1.12 (annotation/tagging wrt consistency) (most in favour, some disagree, significant number of don't know/care)

19:19:13 <DanC> chair notes that we'll get back to ontology version management later in the "general requirements" section

19:19:51 <DanC> metaknowledge: this one is hard for me; I want it, but probably not in the OWL later

19:19:53 <DanC> layer

19:21:13 <nmg> (not under scribe hat) ditto here for reification in general

19:23:11 <JosD> say res1 prop res2 in which res1 and res2 happen to be set of statements (by value!)

19:25:49 <JosD> (by value is either deref res uri or identify by content e.g. N3's { } )

19:26:27 <nmg> jjc: (commenting on 3.4.7) i18n very important

19:26:58 <nmg> ====== Services (Stefan Decker)

19:27:13 <nmg> reqs mailed to www-webont-wg

19:27:18 <nmg>http://lists.w3.org/Archives/Public/www-webont-wg/2002Jan/0083.html

19:27:48 <DanC> nmg, pls do record this discussion

19:28:25 <DanC> Decker: before I start, I'd like to note that a lot is going on in Web Services in a way that's disconnected from Ontologies.

19:28:48 <DanC> ... I suggest we think about how to change that. I don't have much of an idea of how.

19:28:54 <DanC> ... what gets money is XML technology.

19:29:54 <Jah-wowg> ...what gets money is were XML technology focuses?

19:31:55 <nmg> re: efficient inferencing, assumption is made that inferencing will be performed on small devices (rather than on external services called by the devices)

19:33:06 <nmg> fvh: three reqs here: simple and concise lang, layering, defn of limits

19:34:01 <nmg> POLL: language must have a small footprint (mostly against)

19:34:48 <nmg> POLL: language defn is organised in layers (mostly in favour, some against, few neutral)

19:35:03 <nmg> danc: layering not a requirement for this wg

19:35:41 <nmg> POLL: precisely described semantics (jimh overrules vote since this is in the charter)

19:36:13 <DanC> to clarify: from this WG, I don't need more than one layer. I do need layering between this group's language and other groups' langauges.

19:38:05 <nmg> (re: complex types) danc: example given is dates - these are already defined in XML Schema

19:39:06 <nmg> jimh: definition of inequalities (date ranges, eg - see guus' example)

19:40:08 <nmg> danc: XML Schema is a w3c spec, and terms from that spec should be used by this wg in the same way that they are used in XML Schema (complex types)

19:40:18 <nmg> danc: dates (in XML Schema terms) are not complex types

19:41:09 <nmg> POLL: datatypes in language (mostly in favour)

19:41:42 * TimFinin slaps TimFinin around a bit with a large trout

19:42:07 <nmg> danc: requirements should be derived from use cases - this (travel planning) is a good use case

19:43:30 <nmg> jimh: this use case requires range types

19:45:43 <nmg> POLL: ability to express relations between types (eg. inequalities on numbers) (jimh postpones)

19:46:00 <nmg> ACTION on Stefna's group to discuss further tomorrow

19:49:06 <nmg> stefan: (re: unique reference for ontology, class member ship in ontology) need to be able to tell where a class comes from

19:49:58 <nmg> danc: rdfs:isDefinedBy

19:51:56 <nmg> danc: also, ontology defn includes a statement to the effect that it is an ontology defn

19:53:31 <DanC> being able to name ontologies and relate classes to them

19:53:44 <nmg> POLL: ability to give ontologies names and the ability to denote membership of a class in an ontology (mostly in favour, few opposed, few don't know)

19:55:48 <nmg> danc: (re: ontology versioning) ontology as artifact with state

19:56:03 <nmg> jeffh: postpone this discussion to general requirements

19:57:00 <nmg> (re: service/oracle for providing instance data)

19:57:14 <nmg> fvh: also known as procedural attachment

19:57:27 <nmg> danc: existing patent in this area

19:57:32 <DanC> PeterPS: at this point, there's an intellectual property issue. McGuinness and I are authors of a patent relevant to this.

19:57:36 * DanC frowns

19:57:38 <nmg> most details on that, danc?

19:58:04 <nmg> s/most/more/

19:58:43 <nmg> jimh: given IP situation, must et clarification on situation before we go further

19:58:56 <nmg> jimh: chartered as royalty-free

19:59:12 <DeborahMcGuinness-SWWS> some info on patents:

19:59:13 <nmg> ACTION: pfps to determin status of IP on this issue

19:59:13 <DeborahMcGuinness-SWWS> US Patent: 5720008: McGuinness, Patel-Schneider, and Resnick. "Knowledge Base Management System with Dependency Information for Procedural Tests", Issued 2/17/98, Submitted 5/94.

19:59:13 <DeborahMcGuinness-SWWS> US Patent 5974405: McGuinness, Patel-Schneider, and Resnick. "Knowledge Base Management System with Enhanced Explanation of Derived Information", Issued 10/26/99, submitted 5/94.

19:59:19 <nmg> thanks

20:01:08 <nmg> POLL: ignoring IP issue, service for providing instance data (one third each way)

20:03:42 <nmg> pfps: (re: consistency checking of instance data) wrt classification of instance data

20:04:32 <DeborahMcGuinness-SWWS> DeborahMcGuinness-SWWS is now known as DeborahMcGuinness

20:06:36 <nmg> jjc: propose vote on decidability of contstraint checking

20:07:26 <nmg> POLL: Constraints checking: Given a large amount of instance data, it should

20:07:26 <nmg> be possible to check if the instance data confers to a given ontology

20:08:25 <nmg> (split between in favour and don't care, some opposed)

20:08:50 <nmg> jimh: needs to be revisited, further discussion

20:15:44 <nmg> ====== General Requirements (Jeff Heflin)

20:15:54 <nmg>http://lists.w3.org/Archives/Public/www-webont-wg/2002Jan/0062.html

20:22:00 <nmg> jeffh: R1: shared ontologies - already effectively voted on?

20:25:46 <nmg> danc: (re: ontology extension) daml:imports mechanism is worth having. difference between using someone else's terms, and agreeing with all of their defns

20:26:59 <nmg> jimh: explicit mechanism for local vs global imports

20:27:12 <nmg> fvh: large open isse

20:28:05 <nmg> jimh: needs further discussion

20:30:29 <nmg> jeffh: (re: ontology evolution) rdfs recommends using subClassOf for denoting evolution of terms

20:30:47 <nmg> jjc: RDF Schema spec still under discussion

20:31:24 <nmg> ACTION: jeffh to bring implications of this use of subClassOf to attn of RDF Core WG

20:32:19 <nmg> danc: expect this functionality from rules layer, not from ontology layer

20:32:27 <nmg> chair passes to las

20:32:39 <nmg> jimh: this belongs in ontology layer

20:33:34 <nmg> jimh: straightforward starting place for managing versioning in ontologies (paper by jimh and jeffh)

20:33:37 <nmg> pointer to the paper?

20:34:48 <nmg> ontologies reflect social consensus, which changes. ontologies are not static

20:35:08 <nmg> las: take vote on versioning - postponed before

20:36:22 <nmg> POLL: versioning should be requirement for language (mostly in favour, few opposed, few don't care)

20:36:29 <nmg> chair returns to jimh

20:37:36 <nmg> jeffh: (re: ontology interoperability) mapping here is not rule-based

20:37:59 * las hopes we have time for the semantics discussion before she has to go catch her cab....

20:39:21 <nmg> POLL: saubclass/superclass, inverse, equivalence (all in favour)

20:39:52 <nmg> jimh: complex extensions - implication off the table, procedural attachment dealt with elsewhere

20:40:32 <nmg> POLL: complex extensions (mostly against, one in favour, few don't know)

20:41:10 <DanC> note that this poll was in the context of "ontology interoperability" i.e. mapping

20:41:58 <nmg> jimh: (re: R5 detect inconsistency) tabled previously for future discussion

20:42:14 <nmg> danc: need time for future discussion, ideally

20:43:17 <DanC> what I wanted to say is: it's ok to postpone discussion of the decidability requirement for a time, but I hope to get back to it within this ftf meeting

20:43:32 <nmg> sorry - thanks for the clarification

20:45:03 <nmg> fvh: (re: R6 scalability) complexity of reasoning depends on complex DL class defns (as opposed to class use by name in frame systems)

20:45:57 <nmg> (ie. complex defns of unnamed classes)

20:46:11 <nmg> ianh: not necessarily the case

20:46:24 <nmg> jimh: table for future discussion tomorrow

20:48:17 <nmg> jeff: (re: R7 ease of use) feature or design goal?

20:48:30 <nmg> jeffh: no specifics for this

20:48:46 <nmg> jeffh: table XML syntax for future discussion

20:48:47 <DanC> hmm... it does seem useful for our requirements document to say something about the context, i.e.: The Web is Big.

20:49:13 <nmg> jeffh: ditto R9, R10

20:50:58 <nmg> dmcg: (re: C1 explainability) justification for statements in language (cf. proof checking)

20:51:28 <nmg> jjc: (re: C2 i18n) would like this to be a requirement

20:51:44 * DanC realizes he just admitted to not reading all the meeting materials. oops.

20:51:44 <nmg> ====== jeffh finishes

20:53:31 <nmg> jimh: follow-on work for each group - reduce to a couple of use cases, choose writers for such

20:54:21 <nmg> jimh: language features from use case groups

20:54:39 <nmg> jimh: design goals from general requirements group

20:57:37 <nmg> ====== session ends

21:07:06 <heflin> Here's the URL for the paper on versioning:

21:07:09 <heflin>http://www.cse.lehigh.edu/~heflin/pubs/#aaai2000

21:07:36 <heflin> There's also more detail in chapter 3 of my thesis:

21:07:56 <heflin>http://www.cse.lehigh.edu/~heflin/pubs/#heflin-thesis

21:17:41 <nmg> ====== OWL Discussion (Peter Patel-Schneider)

21:20:45 <nmg> tim: (re: representation horror stories) what about the success stories?

21:20:59 <nmg> pfps: relational databases - good theoretical basis

21:21:49 <nmg> pfps: programming languages such as ML

21:24:30 * las is not sure how to be interpreting this presentation. I feel like Peter is stating as fact(oid) things that are simply not.

21:25:07 <las> E.g., there *is* grounding for the web.

21:25:56 <DanC> is there a quick answer to "what's the difference between ZF set theory and flat set theory?"

21:25:56 <nmg> pfps: axiomatisation is a (slight) cheat - grounds out in another proof or model thoery

21:26:17 <nmg> (not-scribe) answers on a postcard

21:26:56 <nmg> pfps: ideally, you have *all* of the above - proof theory, model theory, axiomatisation and operational theory

21:28:18 <nmg> jimh: other formalisations - possible world semantics

21:28:33 <nmg> pfps: possible worlds (kripke) similar to model theory

21:29:53 <nmg> pfps: intuitions re: formalisations not necessarily correct - cf. (early, flat) set theory

21:30:16 <nmg> pfps: zf set theory is believed to be correct

21:30:50 <DanC>http://www-db.research.bell-labs.com/user/pfps/talks/

21:31:13 <sandro> thanks.

21:31:34 * DanC wonders what the problem with DAML+OIL entailment is

21:32:12 * las thinks Peter means that he can't write a semantics that agrees with RDFS semantics and that doesn't have a paradox.

21:32:13 <nmg> pfps: incosistency/paradoxes are one problem, but formalisation may not be right in a more subtle way

21:32:20 <nmg> pfps: ie. daml+oil entailment

21:32:25 <sandro> [ Why does Pat Hayes say RDF has published semantics? Does he view M&S as actually presenting the semantics of RDF? Or does he view his model theory as normative? ]

21:33:24 * DanC isn't at all sure that "we have all seen Russell's paradox"

21:34:54 <nmg> pfps: intuitionist logic rejects 'contradiction implies everything', relevantistic logic (sp?) makes a stronger statement

21:36:43 <sandro> can someone throw in an occastional slide number?

21:36:53 * nmg hopes someone has a camera to capture this diagram

21:37:08 <nmg> slide 13

21:38:11 <nmg> pfps: once commitment is made to syntax and semantics for a representation language, cannot change either

21:38:51 <nmg> pfps: need to decide relationship (in terms of syntax, smenatics and expressive power) between OWL and RDF(S) and XML

21:41:25 <DanC> lynn's point is a good one... relationships between systems include not just "the same" inferences, but S1 sound wr.t. S2 or S1 complete w.r.t S2.

21:42:00 <nmg> las: soundness and completeness refer to relationship between systems

21:42:27 <las> S1 sound wrt S2: If S1 says it's so, S2 agrees. S2 complete wrt S1: If S2 says it's so, S1 agrees.

21:42:46 <DanC> RDFS is complete w.r.t RDF, but not sound.

21:43:03 <nmg> pfps: rdf and rdfs have same syntax and semantics - rdfs complete wrt to rdf, not sound

21:43:26 <DanC> I want OWL to be complete w.r.t. RDFS but not sound.

21:43:54 <nmg> pfps: unsoundness does not mean that you can draw contradictory consequences, only that you can draw more

21:43:54 <las> Specifically, consequential closure of RDF is a subset of RDFS

21:45:00 <nmg> pfps: relationship between rdf and xml - uses xml syntax, ignores xml semantics

21:45:02 <DanC> PFPS: RDF syntax is a subset of XML syntax. RDF semantics are pretty much unrelated to XML semantics.

21:46:36 <las> Oh, also, there's soundness on a subset. E.g., the propositional subset of FOPC is sound wrt propositional calculus, and RDFS's RDF subset is sound wrt RDF (duh)

21:46:39 <nmg> pfps: XML is mapped into a tree (by XML Infoset or XML Query), RDF graph does not correspond

21:46:45 <sandro> "ignores xml semantics" Actually no -- XML has several common semantic forms (see H. Thompson's paper) and RDF/XML basically lets you use each of them, while indicating which you are using so it can all be understood.

21:47:00 <nmg> pointer for the paper?

21:47:05 <sandro> looking

21:47:42 <sandro>http://www.ltg.ed.ac.uk/~ht/normalForms.html

21:48:46 <nmg> pfps: two (different) defns of OWL - OWL, and OWL'

21:48:59 <sandro> But yes -- it is not directly related to the infoset semantics -- it's about the intended semantics of the authors.

21:49:32 <nmg>http://lists.w3.org/Archives/Public/www-webont-wg/2002Jan/0042.html

21:51:23 <nmg> pfps: OWL' much like DAML+OIL, slightly different syntax (slide 15)

21:54:22 <nmg> pfps: anti-foundation avoids some paradoxes, but at a cost (classes not referring to themselves)

21:55:44 <nmg> pfps: model theory based of zf set theory avoids russell's paradox, regardless of foundation axiom

21:56:37 <nmg> url for russell's paradox

21:56:56 <nmg>http://plato.stanford.edu/entries/russell-paradox/

21:58:33 <nmg> pfps: contributing to problems/paradoxes: meta-model (rdf:Class in the model), negation, entailment

21:58:52 <nmg> pfps: removing one of these avoids problem

21:59:02 <jdale> and for the liar's paradox, see: http://www.utm.edu/research/iep/p/par-liar.htm

21:59:03 <nmg> pfps: rdf/s removes negation

21:59:22 <nmg> pfps: daml+oil removes entailment

22:00:13 <nmg> pfps: clarification - daml+oil does not do *rdf* entailment

22:00:16 <Jah-wowg> remove RDF model of entailment

22:01:09 <nmg> jimh: daml+oil has no negation

22:01:18 <nmg> danc: complementOf is a form of negation

22:01:32 <Jah-wowg> no - I said DAML+OIL doesn't have full logical negation (i.e. NOT)

22:01:50 <DanC> russel set in DAML+OIL: "the class of things that have no rdf:types that ...". hmm... missed part of it

22:01:59 <nmg> sorry

22:02:13 <las> pfps: you can construct a Russell-like property by building the class of things that are not the RDF type of anything.

22:02:49 <DanC>http://www-db.research.bell-labs.com/user/pfps/talks/webont-f2f-owl/slide25-0.html

22:03:07 <DanC> check that out, sandro. I'm gonna try to translate to N3 (and maybe to KIF)...

22:03:54 <DanC> I'm quite interested in paraconsistent logics.

22:04:45 <las> Logic families Peter doesn't want to consider: intuitionistic, relativistic, paraconsistent.

22:05:12 <las> Also 3-valued and stratified, but I'm with him on those (although stratified isn't totally out of the question).

22:05:33 <nmg>http://www-db.research.bell-labs.com/user/pfps/talks/webont-f2f-owl/slide24-0.html

22:05:42 <DanC> [[[

22:05:45 <DanC> this <http://www.w3.org/2000/10/swap/log#forSome> :_g1 .

22:05:46 <DanC>

22:05:46 <DanC> :_g1 a :_g1,

22:05:46 <DanC> owl:Restriction;

22:05:46 <DanC> owl:hasClassQ [

22:05:46 <DanC> owl:oneOf [

22:05:48 <DanC> owl:first :_g1;

22:05:50 <DanC> owl:rest owl:nil ] ];

22:05:52 <DanC> owl:maxCardinalityQ "0";

22:05:55 <DanC> owl:onProperty rdf:type .

22:05:57 <DanC> ]]]

22:06:42 <las> This (Does John belong to the intersection of student/not student) is a nice example....it doesn't support reasoning by cases. Really, it's RBC, not LEM (law of the excluded middle) that will cause us the most problem....

22:10:00 <nmg> pfps: classes with reflective defns are not all problematic, but the dividing line between those that are and those that aren't is not simple

22:10:16 <las> RBC=reasoning by cases. (If a, then.... If not a, then....)

22:12:20 * sandro has too-little experience with daml:restrictions to be much use here.

22:12:54 <nmg> pfps: (re: avoiding paradoxes without throwing out one of negation, entailment and the metamodel) adopt an extended syntax for restrictions

22:13:21 <DanC> ==== daml+oil version of russel's paradox, rendered in KIF

22:13:23 <DanC> (prefix-kludge "owl" "http://example/owl-vocab#")

22:13:23 <DanC> (prefix-kludge "rdf" "http://www.w3.org/1999/02/22-rdf-syntax-ns#")

22:13:23 <DanC> (exists (?1_1 ?2_2 ?3_3 )

22:13:23 <DanC> (and

22:13:23 <DanC> (rdf:type ?1_1 owl:Restriction )

22:13:24 <DanC> (owl:onProperty ?1_1 rdf:type )

22:13:26 <DanC> (owl:maxCardinalityQ ?1_1 "0")

22:13:28 <DanC> (owl:hasClassQ ?1_1 ?2_2 )

22:13:30 <DanC> (owl:oneOf ?2_2 ?3_3 )

22:13:32 <DanC> (owl:first ?3_3 ?1_1 )

22:13:33 <nmg> las: tradeoffs in whatever decision we take, but have tradeoffs been discarded out of hand?

22:13:34 <DanC> (owl:rest ?3_3 owl:nil )

22:13:36 <DanC> (rdf:type ?1_1 ?1_1 )

22:13:38 <DanC> )

22:13:40 <DanC> )

22:13:42 <DanC> ====

22:14:04 <nmg> danc: reading pfps' 'we can't' as 'i'd rather not'

22:14:29 <nmg> pfps: not the case

22:15:07 <nmg> dieter: problem arrives when you assume that model theory of owl will be a monotonic extension of that of rdf

22:15:42 <nmg> fvh: price of not writing restrictions as rdf?

22:15:50 <nmg> pfps: can't query *about* restrictions

22:16:01 * las very much regrets that I need to pack up any minute....I really don't want to miss my cab or my flight....

22:16:03 <nmg> pfps: can define classes in terms of restrictions, however

22:16:16 <nmg> pfps: query wrt those

22:16:22 * DanC regrets that too. sigh. logistics are such a bummer.

22:18:07 <nmg> las: socratic approach?

22:19:01 <las> Reference: Jim Crawford (Ben Kuipers), Algernon. Sorry, my browser is closed and I can't find a pointer, but there was a AAAI paper and a UT Austin thesis.

22:19:10 <nmg> thanks

22:20:01 <nmg>http://citeseer.nj.nec.com/crawford91algernon.html

22:20:39 <DeborahMcGuinness> here is one pointer

22:20:39 <DanC> we haven't explored the cost of throwing out (various forms of) negation, to my satisfaction.

22:20:41 <DeborahMcGuinness>http://www.cs.utexas.edu/users/qr/algernon.html

22:20:43 <nmg>http://www.cirl.uoregon.edu/crawford/papers/algy_sigart.ps

22:21:34 <DanC> I find intuitionistic/constructionist logics most promising. i.e. throwing out negation in a way, or throwing out entailment in a way. i.e. making proof harder.

22:22:31 <nmg> las: would socratic querying tamper with entailment

22:22:33 <nmg> pfps: pfps

22:22:36 <nmg> pfps: yes

22:23:55 <nmg> ====== session ends

22:24:07 <DanC> logger, pointer?

22:24:07 <DanC> See http://ilrt.org/discovery/chatlogs/webont/2002-01-14#T22-24-07

22:25:44 <DeborahMcGuinness> and DAML+OIL (March 2001): A Datatype Extension to DAML+OIL (December 2000)

22:25:44 <DeborahMcGuinness> Feedback to www-rdf-logic, please.

22:25:44 <DeborahMcGuinness> DAML+OIL (March 2001) version (revision 4.1): Ian Horrocks, Frank van Harmelen and Peter Patel-Schneider, editors.

22:25:44 <DeborahMcGuinness> The idea behind DAML+OIL (March 2001) is to extend DAML+OIL (December 2000) with arbitrary datatypes from the XML Schema type system (http://www.w3.org/TR/xmlschema-2/#typesystem), while still retaining the desirable properties of the ontology language, in particular its (relative) simplicity and its well defined semantics. This is achieved by maintaining a clear separation between instances of "object" classes (those defined using our ontology languag