W3C Web Ontology Working Group IRC Chat Logs for 2002-10-07

08:59:16 Topic now WebOnt Bristol ftf. scribe: ter Horst

08:59:16 Users on #webont: logger nmg JosD DaveB RRSAgent jjc libby mdean @DanC_jam

09:00:45 <DanC_jam> layering document, from the agenda: "Layering RDFS into OWL" http://www.coginst.uwf.edu/users/phayes/RDFS2OWL-L.html

09:01:56 <DaveB> libby: aha, that's the crucial detail

09:02:16 <libby> yeah, it's not in the help

09:02:30 <DaveB> it is in the man page

09:03:45 <libby> oh, I musta missed it

09:23:45 * jhendler_ would like to hear from some other people

09:48:32 * DaveB waves to jhendler_ , distracting him ;)

10:07:05 <libby>http://swordfish.rdfweb.org/photos/2002/10/07/

10:07:12 <libby> - photos from earlier

10:44:28 <DanC_jam> (resume from break)

10:44:32 <DanC_jam> Mike S. scribing

10:44:42 <DanC_jam> DanC_jam has changed the topic to: WebOnt Bristol ftf. scribe: Mike S.

11:24:34 <DanC> RRSAgent, pointer?

11:24:35 <RRSAgent> See http://www.w3.org/2002/10/07-webont-irc#T11-24-34

11:24:53 <DaveB> logger's also writing it to the web live (at ilrt.org)

11:28:02 <DanC> logger, pointer?

11:28:02 <DanC> See http://ilrt.org/discovery/chatlogs/webont/2002-10-07#T11-28-02

12:33:58 <heflin> Hi everyone, how's the F2F going?

12:54:25 <libby> er...

12:56:37 <heflin> Er?

12:57:47 <libby> :)

12:59:13 <heflin> Are you going to begin the imports discussion in a few minutes?

12:59:46 <libby> sorry jeff, I'm not sure. we're beinging something shortly yes

12:59:57 <nmg> not sure - we were on layering for most of this morning, so I'd guess that we're (slightly) behind sched

13:00:39 <heflin> Layering, huh? Doesn't suprise me that that took a while. Any progress?

13:01:32 <nmg> break-out group is going to thrash out the options; there wasn't any consensus between the large and fast approaches

13:06:32 <heflin> Okay. Could somebody let me know when you get to imports?

13:07:02 <heflin> I'll try to check in here, but e-mail might be best way to contact me (heflin@cse.lehigh.edu)

13:07:26 <libby> ok, we will

13:09:13 <heflin> Thanks!

13:13:12 <libby> no probs

13:19:12 <nmg> jeff, guide discussion has been moved forward (to now)

13:19:58 <nmg> imports and versioning will be at about 15.30 or thereabouts (45mins for guide, 15mins for tea break?)

13:24:31 <heflin> Thanks for the info, Nick.

13:26:27 <DanC> who's scribing now?

13:27:28 <mdean> Mike Dean now scribing

13:27:35 <DanC> DanC has changed the topic to: WebOnt Bristol ftf. scribe: Mike Dean

13:27:43 <mdean> now discussing GUIDE

13:27:57 <mdean> Mike: some changes based on validation, etc.

13:28:05 <mdean> Jim: good feedback from students on history

13:28:42 <mdean> Mike: plan to use bulleted lists for links to previous ontologies and use in previous commercial and major government projects

13:29:34 <mdean> discussion about relationship between xml:base and default namespace

13:29:49 <mdean> Namespaces section

13:30:08 <mdean> refering to Sep 29 version as published

13:30:21 <mdean> DanC: xmlns="#" should be deprecated

13:32:00 <mdean> DanC: not many things currently use xml:base

13:34:10 <mdean> DanC: need to avoid xmlns="#" is due to problems with relative URIrefs and namespaces -- W3C decision not to require this in any of their specs

13:35:17 <mdean> decision to use explicit, non-default namespace with some text discussing alternatives

13:35:41 <mdean> DanC: useful to have some simple "hello world" example

13:36:36 <mdean> Mike wants to modify all examples to fit in wines ontology

13:37:14 <mdean> Jim: this walkthrough is better than previous because it also discusses alternative representations

13:37:53 <mdean> Guus: some desire to make examples more realistic

13:38:09 <mdean> MikeS: make region more accurate geographically

13:38:25 <mdean> Guus: subClassOf example was very part/whole-like

13:38:41 <mdean> Guus: should we include a specific part/whole example?

13:39:06 <mdean> Guus has a part/whole example

13:39:13 * jhendler_ just discovers that those random shootings in the DC area included several in Maryland, including one 2 blocks from my house! glad I didn't know htat...

13:40:21 <mdean> Mike will integrate Guus example

13:40:52 <mdean> Larry Eshelman has how-to example regarding part/whole

13:41:18 <mdean> Jim: feedback is that GUIDE can be more complex

13:41:45 <mdean> Chris: things like this should be in GUIDE

13:41:54 <mdean> Nick: should be in appendix of GUIDE

13:42:42 <mdean> MikeS: want part/whole to be an example of using properties, not how to do part/whole

13:43:50 <mdean> Guus: decided last F2F to have HOW-TO/FAQ separate from GUIDE

13:44:04 <DanC> [I'm getting mixed messages about whether the 'tricks of the trade' is something the WG has decided to do or not]

13:45:05 <mdean> Nick: need to make clear that these are examples of OWL, not an official OWL representation of part/whole

13:46:19 <mdean> Jim: making wine more complete isn't so useful -- want real examples of people things want/need to do

13:46:43 <mdean> Jim: e.g. more examples of complex use of instance data

13:47:33 <mdean> Deb: need example of every language concept

13:47:40 <mdean> Jim: also need more examples of instances

13:48:15 <mdean> Evan: need "HOW-TO" to show how to use OWL for common modelling problems

13:49:19 <mdean> Jeremy: not hearing showstoppers for publishing GUIDE as is -- these may be suggestions for GUIDE v2

13:49:46 <mdean> Guus: agree, current version has received lots of positive feedback

13:50:20 <mdean> Jim: next WD may be last until last call

13:51:51 <mdean> Chris: far more people will read guide than Model Theory, so we need a very good wine ontology, since people start by cut-and-pasting the examples

13:52:06 <DanC> quite; they're not even gonna read the guide. They're just gonna cut/paste.

13:52:21 <mdean> Deb: any particular requests for extensions?

13:52:48 <mdean> Chris: has some suggestions

13:53:26 <mdean> MikeS: perfectly happy with good wine ontology, but not with large/expanding set of appendices (textbook)

13:54:13 <mdean> Jim: maybe we should have examples that go beyond wines ontology

13:54:27 <mdean> MikeS: pulling meals, etc. into separate ontologies

13:54:44 <mdean> MikeS: requires resolution on imports, etc.

13:56:25 <mdean> DanC: good example could change his position to drop imports

13:57:33 <mdean> ~6 people have thoroughly reviewed current draft

13:58:09 <mdean> MikeS: most of current changes can be done pretty rapidly

13:59:39 <DanC> [ooh; I'm interested to know which features he didn't run into/need in this draft.]

13:59:40 <mdean> MikeS: target 2-3 weeks from next Monday

14:00:44 <mdean> MikeD: would prefer use of camelCase rather than LISP-STYLE names

14:01:13 <mdean> Nick: should we commit now to putting the "tricks of the trade" in a separate document?

14:01:56 * DanC q+ to ask if anybody has comments that (a) they haven't sent in email, (b) they feel are important for 1st WD.

14:02:15 <mdean> Enrico: first draft of GUIDE should be correct, need not be complete

14:03:34 <mdean> Deb: should note that example is just one type of part/whole

14:04:29 <mdean> DanC: more cost-effective to get community involved

14:05:44 <DanC> (oh crud; I forgot to ask for pointers-to-issues in the guide, ala in the reference)

14:05:48 <mdean> Jim: target late October for WD, then incorporate larger changes for last call

14:07:15 <DanC> PROPOSED: to publishd the guide, as in http://lists.w3.org/Archives/Public/www-webont-wg/2002Sep/att-0478/01-Guide.html plus edits at the discretion of the editor.

14:08:30 <mdean> MikeS: will note holes/placeholders for missing examples of every feature

14:10:01 <DeborahMc> is there a url where the whole log for this session can be viewed? (I was in the semantics discussion for the beginning sorry)

14:11:23 <mdean> DanC: schedule is quite tight

14:11:58 <libby> Deborah:http://www.ilrt.bris.ac.uk/discovery/chatlogs/webont/2002-10-07.html

14:12:05 <mdean> Chris: some ontology problems need to be fixed before publication

14:12:30 <mdean> Jim: let's discuss technical issues here

14:12:50 <mdean> Stephen: GUIDE may not need to go through formal last call, since it's not normative

14:13:06 <mdean> Jeremy: XML Schema Primer (?) did go through last call

14:13:26 <mdean> Jim: would prefer GUIDE to go through formal last call process

14:13:57 <mdean> John Stanton: what are the real W3C requirements?

14:14:42 <mdean> DanC: alternatives include WD, web page, Note (terminal state)

14:16:07 <mdean> Guus: at F2F 3, HOW-TO document was viewed as a web page

14:17:58 <mdean> Evan suggests withdrawing motions and deferring vote

14:18:09 <mdean> Chris specific comments

14:18:28 <mdean> treatment of regions (part/whole vs. subClassOf)

14:18:40 <mdean> Deb: known problem

14:18:52 <mdean> DanC: transitive, but unrelated to subclass

14:19:09 <mdean> arbitrary point at which classes become instances

14:19:18 <mdean> MikeS: GUIDE notes that this is arbitrary

14:19:32 <mdean> Deb: What is "right" boundary?

14:21:25 <mdean> MikeS: this model is accurate for the domain

14:21:43 <mdean> Guus: perhaps renaming Grape to GrapeType might help

14:21:57 <mdean> MikeS: this is exactly the classes as instances problem

14:22:38 <mdean> MikeS: tried to do some normalization of names

14:23:17 <mdean> Chris: some remaining concerns about relationships between GrapeType and WineGrapeType -- need to think about this more

14:26:05 <mdean> need to note opportunities for subclassing as well

14:26:18 <mdean> Guus: perhaps breakout session on wine ontology

14:28:25 <DanC> ACTION Guus: arrange a break-out session on the Wine ontology and the Guide

14:28:26 * RRSAgent records action 1

14:28:35 <mdean> Jeremy as non-wine expert had trouble following some of this discussion

14:28:57 <mdean> Jim: we aren't content ontology group

14:30:03 <mdean> Guus: time for break

14:30:23 <mdean> breakout group still needs to be scheduled

14:30:58 <mdean> breakout members: Enrico, Chris, Deb, Mike, Evan

14:37:06 <mdean> next session in other room

14:38:04 <Snakker> Snakker is now known as libby_

14:40:18 <libby_> libby_ is now known as libby

14:41:51 <heflin> That's International U.S. by the way...

14:42:23 <libby> whoo-hoo - we got irc port forwarding working

15:02:00 <libby> jeff, we're going to call you shoirtly

15:02:21 <heflin> Great! Thanks, Libby.

15:03:10 <libby> no prob

15:04:20 <mdean_> mdean_ is now known as mdean

15:04:25 <DanC> ping

15:05:15 <mdean> discussing imports and versioning

15:07:12 <libby> libby is now known as libby-scr

15:07:23 <libby-scr> libby-scr is now known as lib-scrib

15:07:49 <lib-scrib> guus: could jeff recap status of imports document?

15:07:57 <lib-scrib> jeff: 3 proposals, 2 similar

15:08:17 * DanC wonders whether my proposal (drop imports) is on the table

15:08:22 <lib-scrib> number 1: outside rdf syntax, inline with rdf spec, concur decision in f2f2

15:08:49 <lib-scrib> advantages: nice suntactic properties (missed somethign)

15:09:10 <lib-scrib> proposal 2: RDF triples. some technical issues in discussion

15:09:30 <DanC> s/concurs/conflicts/

15:09:39 <lib-scrib> ...'undefined' = lack of intereoperability. another document imports something

15:09:59 <lib-scrib> also benfit - existing rdf data can be used as is; this is not tru of 1 - need a translation

15:10:43 <lib-scrib> propsal 3: ? and borden - processing model proposal - syntax like 2, but instead of semantic meaning a processing model

15:10:52 <lib-scrib> ?= massimo?

15:11:27 <lib-scrib> potential problems: leaves things unclear e.g. whether processing occurs before inference...

15:11:41 <lib-scrib> - end of jeff's overview

15:12:29 <lib-scrib> jjc: asks if any of options provide abiltyy to infer triples

15:13:12 <lib-scrib> jeff: inferring triples can lead to complexities, multiple imports. Problem with the thrird option is that doesn;t tell you how to handle inferencing issues

15:13:37 <lib-scrib> guus: comments?

15:13:51 <lib-scrib> danc: getting rid ofg imports is the best option

15:13:59 <lib-scrib> ...having built things

15:14:21 * lib-scrib hopes that's what danc said ;)

15:14:33 <lib-scrib> jeff; dissagrees - ontology extensions important part of requirements doc

15:14:48 <lib-scrib> danc: to extend them doesnt require imports

15:15:20 <lib-scrib> ...each doc has its own meaning; 2 docs meaning is those together...outside the language entirely

15:15:38 <lib-scrib> ...nothing in the language tells you what to import

15:15:47 <lib-scrib> [several people agree w danc]

15:16:38 <lib-scrib> jim: diagrees strongly: users shoudl be able to link to others' onotlogies, change classes etc - orgignal motivation for daml+oil - can't agree but can extend it locally - link back and add their own tersm

15:17:15 <lib-scrib> ...we dont need to be consistent with each other, just with the orginal ontology

15:17:42 <lib-scrib> ??you just tell it where to look for things like you would tell a web browser - the url of teh documnt

15:18:08 <lib-scrib> patH: imports has more meabning than linking baCK

15:18:28 <lib-scrib> jim: use case:

15:18:44 <lib-scrib> class agenda with various properties: chair, issues etc

15:18:58 <DanC> RRSAgent, pointer?

15:18:59 <RRSAgent> See http://www.w3.org/2002/10/07-webont-irc#T15-18-58

15:19:15 <lib-scrib> ...I read it into my tool, producing a form I can fill out, and it goes and get classes for me and does type checking

15:20:02 <lib-scrib> ...I often read in DanC's onoltogy, which has one bit missing - priority. I don;t want to import DanC's document and copy it: it might change - instead I want a live link to DanC's document

15:20:45 * JosD waves to bwm

15:21:05 <lib-scrib> MikeS: this is good software engineering 0- how else do you remmeber this stuff?

15:22:02 <lib-scrib> jim: doesn't cxare about imports as long as can refer to DanC's live document

15:22:15 <lib-scrib> PatH: believe DanC's doc or extend it - different

15:22:43 <lib-scrib> DanC: read Jim's doc most opf teh time - if I need it I get the urls recursively

15:23:20 <heflin> Can I have the floor?

15:23:30 <lib-scrib> e.g. validator - does RDFs closure on a document wrt a vocabulary

15:23:57 <lib-scrib> - implementation experienc that don't need this stuff

15:24:06 <lib-scrib> MikeD: us too: we don;t need imports

15:24:21 <lib-scrib> Jeff would like the floor

15:24:43 <lib-scrib> [crud missed something by jjc]

15:25:58 <lib-scrib> danC: if I use someone elses' property (rather than mention it) (using it rather than annotating it) then I look up the property

15:26:55 <lib-scrib> ...'endorsing' the schema is use it x talliban:y z or fred rdf:typ[e talliban:goodguy

15:27:16 <lib-scrib> jim: we had a conversation that there shoudl notbe an implicit imports

15:27:28 <lib-scrib> ...if we can hacve that, then that's fine

15:27:51 <lib-scrib> patH: jjc said - any use of a voicabulary consitutes can imports?

15:28:49 <lib-scrib> enrico: shows a graph of an application ontology - the arrows mean import, include.

15:29:18 <lib-scrib> ...would like this to be part of the expression of teh languiage. stronger than using the namespace - I substribe to the theory described in the docuennt

15:30:09 <lib-scrib> patH: prob is he doesnt know what it means to not subscribe to an ontology - how can it refuse to import it. you can draw conclusions for it. you cant not assent to rdf

15:31:29 <DanC> they can tell which documents I "import" by looking at what terms I use as properties.

15:31:51 <lib-scrib> jeff: problem with DanC's approach is that un;less there's some list of urls there other people can't work out what documents you need. imports is committing to an ontology - I sort of sanction the infrences made with _these_ documents

15:32:35 <lib-scrib> - important to have 2 kinds of inference - the doc and those that go with iot, to which I agree to - responsibility for what you agree to, versus combining it with anything, to which I haven;t agreed to

15:33:22 <lib-scrib> patH: with no imports mechanism, anyone can combine your doc with otthers and infre things from it - you seem to think a bad thing

15:33:34 <lib-scrib> jeff: no - but require both types, e.g. terms of a contract

15:33:58 <lib-scrib> patH: thinks that buys into a more elaborate theory of teh Seb than we have at the moment

15:34:05 <lib-scrib> s/seb/sweb/

15:35:04 <lib-scrib> MikeS: geog example: import geog vocab; taking only about texas and oklahoma - don;t want to buy into all the recursive onoltogies - jusrt this part of the oklahoma ontology

15:35:26 <lib-scrib> danC: remember only in use though, using the property

15:35:42 <lib-scrib> mikeS getting less nervous, pps more so

15:36:14 <lib-scrib> danc: rdfs:seealso - look at this other document; somethign similar for this? (missed detail, sorry)

15:36:28 <lib-scrib> guus: consensus point - no semantic meaning?

15:36:50 <lib-scrib> patH: from email seealso - often doesn;t point to rdf, often a webpage

15:37:00 <lib-scrib> danc: that's ok - we subproperty it

15:37:42 <lib-scrib> guus: consensus point - subproperty of seealso as a way of docuemnting this imports thing without it having any formal meanning

15:38:03 <lib-scrib> jeff: unacceptable: broken language, because of the usecases - how do you know what you agree to?

15:38:30 <lib-scrib> ...otherwise bascially a centralized knowledge base

15:38:50 <lib-scrib> ...has responded to danC on email -

15:38:53 <lib-scrib> [ref?]

15:39:39 <lib-scrib> jim: you have a difference bwteen imports= pull that document in there vs imports as stronger committment, stronger than DAML

15:40:36 <lib-scrib> ...what about just adding a propoerty to someone's onoltgy? just can;t do it?

15:40:57 <lib-scrib> jeff: maybe by referencing the namespace without importaing it...maybe..?

15:41:11 <lib-scrib> jjc: that seems unlear - danC's position is clear

15:41:16 <lib-scrib> jeeff: not to me

15:41:38 <lib-scrib> mikeS: an operational way of descrbing using tokens from another ontology...

15:41:47 <lib-scrib> danc: no.

15:41:49 <lib-scrib> [scribe lost]

15:42:07 <lib-scrib> danc: if they load the same onologies, they come to same conclusions

15:42:22 <lib-scrib> jim: what if we load diff documents? then diff conclusions

15:42:27 <lib-scrib> danc: not a problem

15:42:45 <lib-scrib> mikeS: but I don;lt know what my assumptions are - I can;t write them down

15:43:22 <lib-scrib> jim: same toolls, same url, same conformance - diff conclusions - problem

15:44:01 <lib-scrib> [big shouty argunment]

15:44:20 <ChrisW> we're shouting so you can hear better, Jeff

15:44:34 <heflin> That's what I thought ;-)

15:44:35 <lib-scrib> danc: need to reference the documents that used to get conclusions

15:44:49 <lib-scrib> patH - these tools are in the wide world, they will pick up things

15:45:07 <lib-scrib> jim at whiteboard

15:45:42 <lib-scrib> pps: disagrees totally with Pat. if yuou import another ontologies, you are require to draw the conclusions

15:45:54 <lib-scrib> patH: no, can;'t require that all conclusions are drawn

15:46:10 <lib-scrib> pps: can;t complain if draw all cnoncclusions

15:46:33 <lib-scrib> pps: doc1 imports doc2. any kb derrived from doc1 entails the stuff in doc2 - is what pps wants

15:46:54 <lib-scrib> patH: must not translate in an operational sense about requiring -

15:47:31 <lib-scrib> enrico: ridiculous to have the semantics depend on the tool - same levle of importance as say subclass

15:47:58 <lib-scrib> ...you are going to subscribe to specofoc onolotgies - a finite number (not like statments)

15:48:23 <lib-scrib> jim: %owl1 -f http://ont1

15:48:29 <lib-scrib> jim: %owl2 -f http://ont1

15:48:36 <lib-scrib> owl1 100% owl compliant

15:48:40 <lib-scrib> owl2 100% owl compliant

15:48:51 <lib-scrib> [ therefore p(x) is undefined]

15:49:03 <lib-scrib> ...that's fine - don;t need to bother with all these semantics!

15:49:29 <lib-scrib> jim v happy with DanC's operational semnatics; need something though, otherwise a crapshoot

15:49:50 <lib-scrib> patH: why would the docs come out sensibly

15:50:04 <lib-scrib> ? thfirst tool has different import idea than second

15:50:17 <lib-scrib> patH: 'importing' is the issue

15:51:00 <lib-scrib> jim: we need a good clean operational semantics

15:52:06 <lib-scrib> danC: this group does not say what every owl implementtaion doesin all languages; the implementation will need to justify its conclusiosn accorting to our spec - it will have certain premises - diffee nt conclusiosn will ahve differenbt set of preferences

15:52:47 <lib-scrib> guus: why are you aginst it?

15:52:59 <lib-scrib> danc: doesnt think will get consensus

15:53:24 <lib-scrib> guus: both agree that have operational semnatics for import and have to say this in the language

15:53:35 <lib-scrib> danc: doesnt always do imports - required it would be a bummer

15:54:00 <lib-scrib> danc: several policies to loda into machine - e.g. uses policy

15:54:30 <lib-scrib> jim: how to we characterise these diferences in policy?

15:55:10 <lib-scrib> danc: we can't use this in general because wants single documents to be well defined.

15:55:28 <lib-scrib> jjc has an example

15:55:30 <DanC> this="implementation defined"

15:55:36 <lib-scrib> jjc: in one file we have

15:55:43 <lib-scrib> <a> z:foo <b>

15:55:47 <lib-scrib> in another file:

15:56:01 <lib-scrib> z:foo domain c

15:56:27 <lib-scrib> c subclassof <<restriction hasvalue v onprooperty w>>

15:56:37 <lib-scrib> does this entail <a> z:w v?

15:56:51 <lib-scrib> jjc thinks jim said that the entailment follws

15:57:02 <lib-scrib> - and agrees

15:57:21 <lib-scrib> ... change of example:

15:57:33 <lib-scrib> <a> z:foo <b> in one doc

15:57:40 <lib-scrib> implicitly importy doc 2:

15:57:55 <lib-scrib> z:foo domain c

15:58:17 <lib-scrib> do these two entail <a> type <c>?

15:58:45 <lib-scrib> danC: not with tools currently,. thopugh you can have a policy which can go and get all the documents andthen entail that.

15:59:01 <lib-scrib> danc: from our spec, the first two triples will entail the other.

15:59:26 <lib-scrib> ? is it allowable for the tool to draw that conclusion form the first tripple?

15:59:45 <lib-scrib> - not within teh spec, although it coudl [...?]

16:01:13 <lib-scrib> ? abase ontolgy, 2 ontologies that extend it incompatibly and monotonically with weach other but not from the original: some tools will find the contradiction, some not

16:01:21 <lib-scrib> DanC: yes - state of the art is this

16:01:46 <lib-scrib> enrico: bizzare to have different conclusions depending on what files it finds

16:01:59 <lib-scrib> patH: not only not bizzare, but the way things are on the web

16:02:14 <lib-scrib> enrico: not what I said [goes to what board]

16:02:23 <lib-scrib> [many converations going on...]

16:03:00 <lib-scrib> onrico: one ontology with 2 stataemnets:

16:03:06 <lib-scrib> man is a subclass of mortal

16:03:10 <lib-scrib> man(socrates)

16:03:20 <lib-scrib> so, mortal(socrates)

16:03:44 <lib-scrib> if the first 2 statements are in two files, I can;t guarantee that I will be able to derrive that socrates is mortal

16:04:11 <lib-scrib> patH: I cant insist that you draw those conclusions, or any onclusionbs at all

16:04:46 <lib-scrib> jjc: in tersm of the spec, we specify the semnatic sfor a complete reasoner, even if none exists.

16:04:56 <lib-scrib> pps: strongly agrees with that

16:05:13 <lib-scrib> jjc: one source of incompleeness will be not importing things

16:05:33 <lib-scrib> danc: cxan we justify the conclusion according to our spec, and we can

16:05:47 <DanC> s/and we can//

16:06:02 <lib-scrib> ok, someone said that

16:06:55 <lib-scrib> jim: what if A:mortal rather than mortal

16:07:03 <lib-scrib> sorry ChrisW

16:07:47 <lib-scrib> jim: if mortal is from a different document, that shoudl not change the conclusion.

16:08:20 <lib-scrib> ...if the other document also said mortal(x) :- foo(x)

16:08:37 <lib-scrib> do I aslo legitimize A:foo?

16:09:01 <lib-scrib> ...mortal subclass transientbeing (from document A)

16:09:38 <lib-scrib> I believe A:mortal socrates; so do I believe socrates is also a A:transientbeing?

16:09:47 <lib-scrib> patH: yes if you got the document back

16:10:36 <lib-scrib> danc: from spec, certain entailmenets follw given certain documents; tool decides the policy

16:10:49 <lib-scrib> chrisW? can you just load onbe property?

16:11:05 <lib-scrib> jim: putting this on our testset

16:11:27 <lib-scrib> patH: entailments are clear - what's the point of the example

16:11:37 <lib-scrib> jim: then you believe that you always importa everything

16:11:58 <lib-scrib> patH: no! you import what you come accross

16:12:06 <lib-scrib> [more shouting]

16:12:54 <lib-scrib> jim: does the premises entail conclusion (from 2 docs)?

16:13:07 <lib-scrib> danc: depends on the premises

16:14:03 <lib-scrib> guus: 5-10 more minutes.

16:14:10 <lib-scrib> ..on this topic

16:14:41 <lib-scrib> sb: entailments are definted in term sof trilpes, not documents - if you have the triples then you ahve the entailments [missed a bit]

16:14:53 <lib-scrib> sb: supports danc, probably

16:15:40 <lib-scrib> chrisW: is theeraany way to include in teh spec some way to tell people when they are required to load a ertain set of triples

16:15:51 <lib-scrib> patH: we can;t require them to in terms of entailments

16:16:10 <lib-scrib> chrisW: youn must load these triples, nothignto do with entailment

16:16:31 <lib-scrib> - i.e. can we say this

16:17:13 <lib-scrib> jjc: manifest file in testcases, pointing to files to include and entailments - get valid entailments _if_ you import the files

16:17:50 <lib-scrib> ...danc saying that can;t use implecit mechanism, but that we can't have imports - we need explcit operational semantics in this language

16:18:57 <lib-scrib> [sorry, scribe can;t keep iup]

16:19:19 <lib-scrib> jim: premise ns:A=http://bla/A

16:19:28 <lib-scrib> man owl:subclass A:mortal

16:19:32 <lib-scrib> socrates a man

16:19:44 <lib-scrib> exists a document http://bla/A

16:20:02 <lib-scrib> which says - mortal owl:subclass :transientbeing

16:20:19 <lib-scrib> conclusioon socrates:a http;//bla/A transientbeing

16:20:42 <lib-scrib> mikeS: software engineering, not entailment test

16:21:09 <lib-scrib> ? this is the kind of thing we should be standardizing

16:21:15 <lib-scrib> patH: yes!

16:21:40 <lib-scrib> chrisW: semantics doesnt tell you what triples to load, just what you get if you do load them

16:22:16 <lib-scrib> ...could do this with every other ttriple

16:22:29 <lib-scrib> patH: you'd have to control what pieces of information people read

16:22:49 <lib-scrib> chrisW: if I pout text on a webpage thenm a nrowser will load it

16:23:05 <lib-scrib> danc: images? - commonality of implementation, not a standard

16:23:13 <lib-scrib> ...ok with me as long as I don;t have to use it

16:24:10 <lib-scrib> [....]

16:24:53 <lib-scrib> danC; has lots of ontloghies that has made on a daily basis, with coauthors, also mikeD, jos. we could like with no particular policy

16:25:43 <lib-scrib> ian: if my reasoner can load anythign at all from the web, all we need for a fillly comlpiant ontology is somethign that says yes to every entailment and ...do something...sorry?

16:25:55 <lib-scrib> patH: we can;'t pass laws against stupidity

16:26:07 <lib-scrib> mikeS": we can try to help them

16:26:26 <lib-scrib> patH: we can;t heplp them because we don;t know what they will do with it

16:27:22 <lib-scrib> jim: we will come up with agrpah that only contains certain facts, that's consistent. we dont feel comfortable coming up wityh a solution wrt imports - we think the following policy looks good but we dont wantto endorse it as a standard

16:27:31 <lib-scrib> several peopel can live with it, ? can;t

16:27:58 <lib-scrib> ? wants to state a policy - pick a policy

16:28:21 <lib-scrib> ? picking one is a standardization issue

16:28:34 <lib-scrib> danc: we dont have enough informationt opick tyjhe right one

16:28:45 <lib-scrib> guus: should we pick one even if it's the wrong one?

16:28:56 <lib-scrib> ?what does it mean to be riht? we can add a policy later

16:28:58 <nmg> ? == EVAN WALLACE

16:29:05 <nmg> sorry, caps lock

16:29:07 <lib-scrib> (thanks, sorry)

16:29:45 <lib-scrib> chrisW: a proposal: if I write an ontology, want some assurance that it will be used consistently. some peopel say can;t do this on the web, don;t want this etc...

16:30:16 <lib-scrib> ....but if you could have an optional tag to say - loda all these triples between the tags - for owl complience. don;t have to use it

16:30:46 <lib-scrib> ...a packaging mechanism

16:32:08 <lib-scrib> patH: doesnt think that chris's proposal meams anything. whta can they do or can;t do if someone else reads this packaging

16:33:10 <lib-scrib> chrisW: sees a need to package together owl assertions - a compliant tool cannot pick and chosse form the triples

16:33:30 <lib-scrib> patH; so a tool can't use a part of teh ontology?

16:33:42 <lib-scrib> chris: just loading it

16:33:58 <lib-scrib> patH: so all it does is say that I have to juse up more of my memory!

16:34:40 <lib-scrib> danC: maybe we shoudl ahve an issue about tool conformance - we dont yet

16:35:31 <lib-scrib> ian: our spec tells you whether this wentailments follws form these triples; if you get more info from the web, then it might entail when it shoudl not

16:35:44 <lib-scrib> danC: different premises!

16:36:01 <lib-scrib> ian: more premises, more conclusions - and he's entitled to do this

16:36:13 <lib-scrib> danC: not more conclusions, different premises

16:36:24 <lib-scrib> ian: no way of specifying the premises!

16:36:50 <lib-scrib> jim: rss - like ontology 'channel' proposal

16:37:49 <lib-scrib> jim: <owl:onoltogy rdf:id>

16:38:01 <lib-scrib> <owl:includes rdf:collection..?

16:38:09 <lib-scrib> <ontlist:uri1>

16:38:11 <lib-scrib> <ontlist:uri2>

16:38:14 <lib-scrib> <ontlist:uri3>

16:38:27 <lib-scrib> ...if you point at me I will expect you to read in all these uris

16:39:05 <lib-scrib> ...could be just a policy; could be a requiremenet [scribe - what's the difefrence?]

16:39:51 <lib-scrib> patH: clarification q: suppose in order to conform I import the whole thing. suppose I then use a tiny bit I draw a conclusion and then throw it - why did I need to ijport it all?

16:40:29 <lib-scrib> jim: chris can be assured that if you change something from within the imports.....[missed stuff]

16:40:38 <lib-scrib> ...an explicit way to grow the graph

16:40:59 <lib-scrib> jjc: this example looks like the begiining of something - needs to be worked out

16:41:33 <lib-scrib> ...an incomplete reasoner can just look at some part sof some documents - no problem

16:41:58 <lib-scrib> ...nonl;y prob is when you say tghat scrates is not a transient being

16:42:14 <lib-scrib> ian: you are entitleed to use up to the fuull set of docs that are specified

16:42:44 <lib-scrib> patH: you are alsways compplete with repect to a particular knowledge base

16:43:37 <lib-scrib> ian: tehre shoufdl be a list to show the list of triple you are intended to conclude

16:44:21 <lib-scrib> patH: we will get a supergoogle

16:44:22 <lib-scrib> ...

16:44:30 <lib-scrib> [jeff, guus says sorry]

16:44:35 <lib-scrib> [scribe loses track

16:44:36 <lib-scrib> ]

16:44:41 <heflin> Sorry for what?

16:45:03 <lib-scrib> [for the lack of clearness maybe?]

16:45:59 <lib-scrib> guus: we're going in circles; 3 positions

16:46:05 <lib-scrib> [scribe didnt catch them!]

16:46:33 <lib-scrib> [did anyone catch them?]

16:47:10 <lib-scrib> chrisW: a way to convey the intention of the author of the ontology is maybe what trying to get at

16:47:38 <lib-scrib> guus: shoudl these principles be pragmatic, formal or not bother

16:47:56 <lib-scrib> danC; 2 diff types of testcases, one that follow links, one that doesn't

16:48:17 <lib-scrib> ...call it a 'slurp it up and look for entailments tests' or something

16:48:37 <lib-scrib> ...give it the same status as the rest of our tests

16:49:29 <lib-scrib> jim has a strong problem with that one. import doesnt tell you what to import except the entire document, prob with metdata, versioning

16:49:45 <lib-scrib> ...is it a graph ort an xml document that inetrested in?

16:49:54 <lib-scrib> ...get a small group to do a compromise position

16:50:14 <lib-scrib> ...being ab;le to name a set of statemenets together, then you culd name an ontology; rather than a docuemnt

16:50:37 <lib-scrib> mikeD: confused - sound like imports ok with danC now

16:50:51 <lib-scrib> danc: ok with it as long as specified separately from entailments

16:51:22 <lib-scrib> jjc: if difefent sorts of entailment tests, it shoudl be interms of the way that they are incomplete

16:52:00 <lib-scrib> ...an extra layer of metadata over teh tests to specify the soret of reasoning you need for them. one of lots of ways of being incomplete - not a special one

16:52:19 <lib-scrib> danc: not incompleetness, extra-logical but maybe we can agree using testcasees

16:52:35 <lib-scrib> ....importing less than a whole doc - any impleemntations?

16:52:48 <lib-scrib> jim: yes - get it all, but just get a few uris

16:53:43 <lib-scrib> jos: would be good if we could come up with a proof of the conclusion - explain why, a justification, evidence

16:54:25 <lib-scrib> action connolly to raise and postpone a justification mechanism

16:54:45 <lib-scrib> jim: - on an issues list for teh future

16:54:55 <lib-scrib> - useful - a non-trivial issue

16:55:25 <lib-scrib> guus: proposals...?

16:55:49 <lib-scrib> danc: informally, anc's proposal is consistent with charter and jeff's first proposal

16:55:53 <lib-scrib> [missed what was]

16:56:19 <lib-scrib> danc thinks jjc has a few propsed tests coudl discuss at a later telecon

16:56:47 <lib-scrib> patH - we could say that that's part of the definition of entailment

16:57:12 <lib-scrib> danC; want a separate definition of a triple taht doesn;t depend on the sdtate of the world [..?]

16:57:48 <lib-scrib> jim: the real issue is how are we going to write imports down, if imports has a daml like meaning (does imports go in teh graph?)

16:58:19 <lib-scrib> guus: at amsterdam f2f, exchange syntax is rdf/xml, and so only proposal 2 is acceptable

16:58:28 <lib-scrib> ? proposal 3?

16:58:46 <lib-scrib> danc: from syntax 2 and 3 are not distinguishable

16:59:04 <lib-scrib> jim: use syntax of 2 and the implication mechanism of 3 - assume a strong imports

16:59:08 <lib-scrib> guus: and have test suites

16:59:53 <lib-scrib> action Danc - make sure the test cases from today are written down

17:01:23 <lib-scrib> proposed resolution - `we will usew the syntax of `jeff's `proposal 2

17:01:38 <DanC> PROPOSED: to include daml:imports syntax as in the reference WD, to include an operational specification of how to find the triples before testing semantic entailment. To have a separate class of tests for this operation.

17:03:02 <DanC> jeff, you ok with that?

17:03:22 <lib-scrib> scribe has to go....

17:03:28 * DanC accepts scribe duties

17:04:07 <lib-scrib> thanks danc

17:05:53 <DanC> PROPOSED: to include daml:imports syntax as in the reference WD, to provide a specification of the set of triples to be included in the test of semantic entailment. To have a separate class of tests for this operation.

17:06:19 <DanC> straw poll...

17:06:24 <DanC> MikeD doesn't like it

17:07:16 <DanC> MikeD: daml:imports adds a lot of hair... I don't think we've clarified why it's better than the mechanism Dan talked about [i.e. if you use a property, ...]

17:08:41 <DanC> [... question of whether daml:imports is sufficiently valualbe, the cost of implentation, the obligation or lack thereof of implementing it]

17:09:18 <DanC> Volz: I don't like this because if we have imports, we need a "if that fails..." mechanism.

17:10:22 <DanC> Guus: perhaps we can postpone the decision until our 'recap' session tomorrow.

17:10:49 <DanC> [Guus notes nobody joined 'the telcon']

17:10:56 <DanC> thanks for joining, jeff.

17:11:02 <DanC> ADJOURN

18:31:35 <heflin> quit

20:32:55 * RRSAgent excuses himself; his presence no longer seems to be needed