W3C

- DRAFT -

RIF Telecon 24 Apr 07

24 Apr 2007

Agenda

See also: IRC log

Attendees

Present
Allen Ginsberg, Chris Welty (ChrisW), Christian de Sainte Marie (PhilippeB), Deborah Nichols, Gary Hallmark, Harold Boley (Harold), Hassan Ait-Kaci, Jos de Bruijn (josb), Leora Morgenstern, Michael Kifer, Paula-Lavinia Patranjan (PaulaP), Peter F. Patel-Schneider (pfps), Philippe Bonnard (PhilippeB), Sandro Hawke (sandro), Stella Mitchell
Regrets
Dave Reynolds, David Hirtle (at a conference), Franšois Bry, Igor Mozetic, Markus Kr÷tzsch, Michael Sintek (project meeting)
Chair
Chris Welty
Scribe
Stella Mitchell

Contents


Admin

ChrisW: Agenda ammendment: we will also discuss the next F2F today

ChrisW: Next telecon, Tuesday May 1st

sandro: May 1st is a holiday for some

ChrisW: How many people who are here today can attend next week? Please indicate in the IRC log.

<sandro> +1 will attend next tuesday

<josb> +1

<LeoraMorgenstern> +1 for attending next Tue.

<Harold> +1

<Hassan> +1 will attend may 1

<Deborah_Nichols> +1

<pfps> +1 for next week

<AllenGinsberg> +1

ChrisW: 8 people said yes...we will hold the telecon next week

<PhilippeB> +1

ChrisW: The week after that, May 8, we may have to cancel due to a W3C conflict. We will discuss by email.

F2F

ChrisW: The next F2F is in just over a month, in Innsbruck, Austria

JosB: Preparations are on schedule; more information is on the wiki. I will need information from the WG about the number of attendees for the meeting and the number of attendees for dinner on June 2nd.

ChrisW: Sandro, can you set up a questionnaire, and set a deadline of May 17th for answers?

sandro: Yes, I will. Also, note that some people may attend on one day and not the other, so we should capture that on the questionnaire.

<Hassan> lunches?

JosB: The meeting will be at the university at DERI

ChrisW: What is the best hotel, to be close to meeting?

JosB: I think there are none in the vicinity, but I will double check. People usually stay in the center, and there is a good bus connection from there to the meeting location

ChrisW: Add any information to the wiki that will be helpful, such as what hotels are close to the bus, etc

<Hassan> any local info will be good

sandro: We should encourage people to make travel plans at the same time they fill out the survey so that they don't forget. There should be enough information on the wiki to allow them to do that.

JosB: The "Getting there" page is redundant with "Reaching Innsbruck" page. I will update the wiki

ChrisW: Any other questions for Jos? ... (none)

<sandro> ACTION: Sandro to make F2F6 attendance survey [recorded in http://www.w3.org/2007/04/24-rif-minutes.html#action01]

<rifbot> Created ACTION-278 - Make F2F6 attendance survey [on Sandro Hawke - due 2007-05-01].

Liason

ChrisW: Any news from liasons?...none are on the call

RIF Core

ChrisW: I didn't see any public comments on the core WD this week

<PhilippeB> none

sandro: The last comment was April 3rd

RIF Core WD1 outreach actions

sandro: Did anyone send out any reminders for relevant people to review RIF Core WD1?

sandro: No one has advertised it except for me

<sandro> Outreach page on wiki

<Deborah_Nichols> I sent to some folks at SRI but forgot to update the contacts page - will do after the meeting. I told them we would accept comments after the deadline

ChrisW: The deadline for comments is April 27th

Christian: I intended to notify some groups, but didn't do it yet. Can we extend the comment deadline?

sandro: Yes. We can't change the deadline date printed on the document, but we can let people know that we will still accept comments after the deadline has passed.

<sandro> Sandro: "Even though the deadline is passing, we will still accept comments"

ChrisW: Christian, you did javarules last time?

<PhilippeB> you can give me javarules, omg

<sandro> ACTION: Christopher announce rif-core to CL [recorded in http://www.w3.org/2007/04/24-rif-minutes.html#action02]

<rifbot> Created ACTION-279 - Announce rif-core to CL [on Christopher Welty - due 2007-05-01].

<sandro> ACTION: christian announce rif-core to Java Rules [recorded in http://www.w3.org/2007/04/24-rif-minutes.html#action04]

<rifbot> Created ACTION-280 - Announce rif-core to Java Rules [on Christian de Sainte Marie - due 2007-05-01].

ChrisW: Any volunteers to announce core WD to the business rules community?

<PhilippeB> give it to csma

<sandro> ACTION: Christian account rif-core to BRCommunity [recorded in http://www.w3.org/2007/04/24-rif-minutes.html#action05]

<rifbot> Created ACTION-281 - Account rif-core to BRCommunity [on Christian de Sainte Marie - due 2007-05-01].

ChrisW: I will do OMG. Also, I already notified Lee Feigenbaum (chair of the DAWG working group) that we are looking for comments

<sandro> ACTION: Christopher announce rif-core to OMG [recorded in http://www.w3.org/2007/04/24-rif-minutes.html#action06]

<rifbot> Created ACTION-282 - Announce rif-core to OMG [on Christopher Welty - due 2007-05-01].

ChrisW: MISMO brew - do we need to notify them again?

<PhilippeB> OK

<sandro> ACTION: Christian to accounce rif-core to MISMO [recorded in http://www.w3.org/2007/04/24-rif-minutes.html#action07]

<rifbot> Created ACTION-283 - Accounce rif-core to MISMO [on Christian de Sainte Marie - due 2007-05-01].

ChrisW: Please everyone, spread the word around to any relevant groups or people that you are in contact with.

test Raindance infrastructure in preparation for online demo during future telecon

ChrisW: We will now try the online demo setup. We want to be able to share RIF demos in future telecons

<PhilippeB> http://ilog.on.raindance.com

<Hassan> I installed it - then what?

Christian: Go to "join unscheduled meeting"...then choose level of download

<PhilippeB> Conference ID: 9250520

<PhilippeB> https://ilog.on.raindance.com

ChrisW: Will this allow each of us to give a demo? We don't want to have to download different software every time a different person wants to give a demo.

<Harold> Should we use the Full Version or the Light Version for the purpose of the demo today?

Christian: Look for 'meeting room'

ChrisW: It's working for me... Hassan, were you able to find it?

<AllenGinsberg> Hassan: open the application and you can join that way

sandro: Can Hassan and Phillipe debug Hassan's problem after the meeting?

<PhilippeB> Yes, I will help anybody if needed after the meeting

ChrisW: Did anyone else try and fail?

Deborah_Nichols: I got an error when I downloaded the full version, I am now trying the light version

<GaryHallmark> If one of us "requests control", can we then present info from our computer to the rest?

ChrisW: Hassan joined successfully. Let's give another minute for questions/problems
...it seems to be basically working

<PhilippeB> OK

ChrisW: When do you plan to do the demo?

PhilippeB: We will plan to do it next week, even though it will be May 1st and attendance may be light.

Hassan: How long will the demo take?

PhilippeB: Between 10 and 15 minutes

Issue 25 (RDF in RIF Core) - sub-issue of BNodes

ChrisW: There hasn't been much email about this in the past week. Sandro, can you give a quick summary of the BNodes issue?

<ChrisW> Sandro's message: http://lists.w3.org/Archives/Public/public-rif-wg/2007Apr/0029.html

sandro: BNodes in RDF are logically equivalent to existentially quantified variables with document scope
... question is how to map this to RIF?
... we need to figure out how other rule systems use BNodes and how that would map to RIF core
... I think RIF core already has what we need to handle this
... RIF core has file scope existential variables (i.e. the local names), even though they are not called that
... this is my thinking so far, although I haven't spent a lot of time on it, so I don't have an extremely high level of confidence in my conclusions

ChrisW: A BNode is an implicitly existentially qualified variable, but a local name in RIF is not that. You would need quantification also.

sandro: That depends on the RDF graph...
...btw, I got it wrong in my email.

JosB: There is a semantic difference:
... you could use globally unique RIF local names to represent RDF BNodes, but it will not emulate everything in RDF, for example, during the merging of RDF graphs
... two different BNodes can indicate the same thing, but two different local names will typically indicate different things

sandro: That is not necessarily true

<josb> p(_:a)

<josb> p(_:b)

JosB: If you merge the above two in RDF, you will know that you have one thing, but under your proposal you will know that you have two things

Hassan: Why isn't there a notion of a variable in RDF? i.e. a local name that can be bound and dereferenced

sandro: There has been lots of discussion behind the decision

ChrisW: Besides, it's an established standard and we have to understand how to be compatible with it

sandro: I think we can think in terms of variables

JosB: We don't have a notion of existentially quantified variables in facts in RIF

<Harold> Sandro, RIF Core cannot currently assert 'existentially connected' facts as in Exists ?x (And ( p(a ?x) q(b ?x) ) ), while BNodes can express such 'document-global' existentials.

Harold: I put an example in the IRC above

sandro: What if you have a contstant name? how is that different from an existentially quantified variable?

JosB: There is a difference: with a skolem constant you have a name, with variable you don't have a name
... I don't have a good example at the moment, I will think about it

JosB: For certain things Sandro's proposal will work (e.g. entailment), and for others it will not work (e.g. merging)

ChrisW: Jos will come up with an example

<sandro> ACTION: JosDeBruin to present test case where b-nodes don't directly map to file-scope constants [recorded in http://www.w3.org/2007/04/24-rif-minutes.html#action09]

<rifbot> Sorry, couldn't find user - JosDeBruin

<sandro> (I put the action in tracker via the web interface.)

<josb> In fulfillment of my action about bNodes: http://lists.w3.org/Archives/Public/public-rif-wg/2007Apr/0062.html

Harold:Replacing existentials with skolem constants should work if we 'localize' all constants to the documents in which they occur

Harold: If a constant "John" occurs in a document, it could be made unique by using the document's url as a unique prefix as in "http://example.org#John"/

<sandro> Harold, then it's not a file-scope name, it's a relative URI.

<sandro> (which isn't good enough for some applications, as I recall.)

ChrisW: Any other discussion on this topic?

MichaelK: I think there's a problem with this when you are trying to count
... if you skolemize it will be a separate constant...and then if you want to count the number of nodes....
... for example, say you have 5 normal constants and 10 bnodes. If you skolemize and then count, your result will be 15 nodes, but if you don't skolemize, your result will be 5 nodes

ChrisW: I'm not sure we need to be able to count. Do we have equality in the core?

MichaelK: Yes, we do have equality.
... when we introduce aggregate functions maybe we will have problems
... if we treat them as Sandro suggested, they will be treated as different things

sandro: Only if you have a unique name assumption

ChrisW: An existentially quantifed variable is not "equal" to something else, rather you can bind it to something

JosB: You can formulate a question: say, are there different x and y so that b(x) and b(y) hold?
... as soon as you start counting your query answers, the difference will show up
... yes, SPARQL returns BNodes

sandro:That is bad practice

ChrisW: Any other points? ... (none)

UCR

AllenGinsberg: I have an outstanding action-274 that I need to work on - it is continued

RIFRAF

ChrisW: I thought Axel would be here, but he isn't

Leora: This morning I sent an email about my action-173. I built an ontology on top of what Allen had done. I also described some issues that I identified during the exercise

<sandro> Leora's e-mail

Leora: The original RAF had discriminators for ECA rules
...ECA is generally for updating DBs
...I wanted to expand it to cover action languages in general
...an action language would be useful, for example, in the medical decision support use case
...these action languages all have some features in common, but there are also differences between them
...the interesting exercise was to figure out how to categorize the features. For example, does the distinction belong to the model or the theory?
...we can discuss my results

sandro: Where are we as far as ontologizing RIF core?

Leora: I did this based on use cases to see what we would need in terms of RIF core syntax. The action rules can be represented as rules with heads and bodies
... I classified these items under categories that we already have in the hierarchy, where possible

sandro: Is there a way with protoge to show the diagram, so that we can see it all at a glance?

ChrisW: Go to the project menu, choose configure, turn on OWLviz and that plugin will draw a picture
... you may need to install a graph package

sandro: If someone can manage to get a picture of this, please mail it to the RIFWG list and say how you created it

ChrisW: It will be a big picture

AllenGinsberg: Axel started with Sandro's OWL file, I started with Axel's, and Leora started with mine, so we have all built on each other

<Hassan> The rest of us who do not know Protege need time to familiarize ourselves with it off line

Leora: I don't understand why conjunction is under disjunction in the ontology that I started with.

AllenGinsberg: It looks like conjunction was accidentally moved under disjunction by you Leora. It is not like that in either Axel's or my version.

ChrisW: Axel was trying to classify rule language features, i.e. the things you would find in the syntax of a conventional rule language
... you (Leora) are extending it with additional things, such as temporal concepts

Leora: I think the medical decision support use case needs a relatively expressive action language, something like event calculus or situation calculus
... I think these extensions will be needed to cover the use cases at some point

PaulaP: RIFRAF is about rule languages and systems, not about concepts So I think we should be looking at languages while making the discriminators/ontology, and not at use cases

Leora: I was looking at actual languages

ChrisW: I would like to suggest a methodological approach that will allow us to distinguish different parts of the ontology.
... OWL has an import feature. We can have an OWL version of the core features and then people can import that file and extend it
... So we can have Axel's piece as a separate file/feature that can be imported, and then you can add your work as an extension on that, and save it in a separate file.

ChrisW: Leora, these features you covered look like they are far in the future for RIF. We will not have an explicit notion of time in the rule language in the short term.

Leora: So you are not expecting to be able to translate into subsets of e.g. event calculus?

ChrisW: We would need a shared theory of time
... we don't anticipate handling these features by the end of this first phase of the WG, i.e. by November.

Leora: So I should ontologize a simple subset of one language such as \cal{A}?

ChrisW: Sure, but use Axel's ontology as a base, import it, extend it, and save it in a separate file.

ChrisW: Allen, would you like to summarize what you did with your section?

Allen's e-mail

AllenGinsberg: I tried to represent the questions/discriminators in section 3 (Semantic Discriminators)
... as an example, I had 'semantic structure' as a class
... I tried to add properties and restrictions
... also, inference procedures - whether they are decidable or undecidable
... I added comments indicating what I was trying to capture. But you would really need a rule language associated with this ontology.
... I didn't cover the part about modal operators yet

ChrisW: How much more do you have to add?

AllenGinsberg: Just the part about modal operators/higher order logics. This was difficult to deal with.

ChrisW: OK, each of you move your pieces into a separate OWL file that imports Axel's base

ChrisW: Paula, you were originally part of this effort. Do you want to add something here?

PaulaP: I'm not sure. We looked at five reactive rule languages and we derived the original ECA discrimminators from those.
... I'm not sure if my approach is compatible with leora's
... I will look at Leora's work and decide on the next step

<Hassan> Paula is (like me) going BU while they go TD

ChrisW: Any other comments? ... (none)

AOB

ChrisW: Any other business? ... (none)

ChrisW: I propose to adjourn

<Hassan> +1

<PaulaP> +1

<sandro> +1 adjourn

<PaulaP> bye

Summary of Action Items

[NEW] ACTION: Christian account rif-core to BRCommunity [recorded in http://www.w3.org/2007/04/24-rif-minutes.html#action05]
[NEW] ACTION: christian announce rif-core to Java Rules [recorded in http://www.w3.org/2007/04/24-rif-minutes.html#action04]
[NEW] ACTION: Christian to accounce rif-core to MISMO [recorded in http://www.w3.org/2007/04/24-rif-minutes.html#action07]
[NEW] ACTION: Christopher announce rif-core to CL [recorded in http://www.w3.org/2007/04/24-rif-minutes.html#action02]
[NEW] ACTION: Christopher announce rif-core to OMG [recorded in http://www.w3.org/2007/04/24-rif-minutes.html#action06]
[NEW] ACTION: csma announce rif-core to Java Rules [recorded in http://www.w3.org/2007/04/24-rif-minutes.html#action03]
[NEW] ACTION: Jos to present test case where b-nodes don't directly map to file-scope constants [recorded in http://www.w3.org/2007/04/24-rif-minutes.html#action08]
[NEW] ACTION: JosDeBruin to present test case where b-nodes don't directly map to file-scope constants [recorded in http://www.w3.org/2007/04/24-rif-minutes.html#action09]
[NEW] ACTION: Sandro to make F2F6 attendance survey [recorded in http://www.w3.org/2007/04/24-rif-minutes.html#action01]
 
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.128 (CVS log)
$Date: 2007/04/24 17:12:45 $
--=_mixed 0065C7EB852572C9_=--