RIF F2F2, day 2 -- 28 Feb 2006

<JeffP> Mar 8 will be the new deadline of the UCR document

<JeffP> then WG review and we will vote on 14 Mar

<JeffP> just the UC section

<JeffP> questions?

<JeffP> Now get to the use cases ...

<JeffP> 2.7

<JeffP> Third Party Rule-Interchange Services

<JeffP> Marla's comment is the same

<JeffP> Any discussion of the use case?

<JeffP> csma: some concern similar to the 2.5 Human-oriented Business Rules

<JeffP> Allen: it could be, but this is natural uase case for ...

<josb> ScribeNick: JeffP

but I am not sure it is similar

any device can have mistake, ... I don't know whether we should label it as compliant

csma: then seems that they are different use cases

Allen: it is a very broad area and RIF will be very useful

Chris: any more comments?

2.7 is approved

Now 2.8 Rich Knowledge Representation

Chris: marla has the same comment?

discussion ...

Sandro: I don't understand the section, I see only OWL ++

csma: My understanding is it is not only extension of OWL ...

the point is the rules are needed to be part of the vocabulary

Paul: are there any existing rule extension of OWL?

Chris: there are

csma: I used OWL as a justification

FranÃ§ois: My suggestion is to make the RIF language extremely expressive

csma: does it make sense to put the uncle example with the brain one

Christin: the brain one is a real example and the uncle one is not

Sandro: I think we need a real one

Gavin: we need some text to show the connection between OWL and RIF

<scribe> ACTION: FranÃ§ois to contact Ian about the change of the use case [recorded in http://www.w3.org/2006/02/28-rif-minutes.html#action01]

Chris: Michael, can you explain your comments?

Michael: I think the big picture of the section is fine but we should make it simple for readers to read

Chris: do you have any specific comments on change?

<scribe> ACTION: Michael to provide some specific comments of 2.8 [recorded in http://www.w3.org/2006/02/28-rif-minutes.html#action02]

Michael: maybe we should extend the use case and make it more general

Sandro: ...

Chris: [summarise what Sandro said] We should change the name, existence of OWL is a justification of the need of exchange ontologies

Proposed new name: Interchanging extended ontologies

because there are no standard to cover that

FranÃ§ois: it seems too technical and I would like to see the applicataions

scribe: in the title

A main title about the application field, and a technical sub-title

Sandro: sounds like a good idea

Christin: I think the application title is too narrow, what about "compatability between ontologies and rules"

this is required by many use cases in many different field

Sandro: OK I see the point

Sandro propose: Interchaning rule extensions to ontology languages

Chris: do we need to consider languages outside W3C standards?

any languages?

csma: topic map

pfps: CYC-L

csma: common logic too, although it is not an ontology language and more like a rule language

Sandro: what aout "Interchaning rule extensions to RDF and OWL"

<pfps> USE CASE 2.8: Rich KR

<pfps> Don: What about mismatches between source and target systems?

<pfps> Chris: That is a different use case.

<pfps> Chris: Title: Interchanging Rule Extensions to OWL (for Brain Anatomy Research)

<pfps> Rationale: That is the use case.

<pfps> ?: Does this cover RDF?

<pfps> Chris: No, should it be covered?

<pfps> Jeremy: No, this use case is adequate (for HP).

<pfps> ?: This loses the idea of rich ontology languages.

<pfps> Chris: This would be a new use case.

<pfps> Michael: Leave the qualifier off the title, to allow for other examples.

<pfps> Francois: The qualifier is a good title - it says it all.

<pfps> Titles need to be concrete.

<pfps> Chris: Objections to the title.

<pfps> Giorgos Stamou: The use is general, the example is just illustrative.

<pfps> Chris: Title options 1/ no clause 2/ "for" clause 3/ subtitle

<pfps> Chris: Title: Interchanging Rule Extensions to OWL -

<pfps> A Brain Anatomy Research Example

<pfps> Chris: Actions 1/ remove uncle example and replace with brain anatomy example

<pfps> 2/ make title be Interchanging rule extensions to OWL

<pfps> ACTION on PFPS - ask Ian to make changes to use case

<pfps> Chris and Michael: modify Michael's action to think about new, related use case

<pfps> USE CASE 2.1: Information Integration

<pfps> Sandro: Right thing is to have Frank revise to address concerns

<pfps> ?: Perhaps the right thing is to have 2 use cases

<pfps> Christian: New topic is "process and supply chain"

<pfps> ?: Perhaps might integrate remaining part of 2.6

<pfps> Sandro: New title is "access to business rules of supply chain partners"

<pfps> Frank: 2.6 example is FOAF-style rules about telephone contact rules

<pfps> Christian: Proposal 1: remaining part of 2.6 fits into 2.1, so 2.6 is deleted

<pfps> Proposal 2: remove data integration from 2.1, retitle, and revise,

<pfps> ensuring that there are rules there

<pfps> Proposal 3: add new use case about data integration

<pfps> Sandro: moving 2.6 to 2.1 doesn't make sense

<pfps> Christian: ok

<pfps> Chris: objections to Proposal 2?

<pfps> ACTION on Frank - effect Proposal 2 above

<pfps> Chris: discussion on Proposal 3

<pfps> Sandro: I can take the stuff from the charter and make it into use case 2.9

<pfps> Title: Vocabulary mapping for data integration

<pfps> Christian: consult with Dave Reynolds

<pfps> Sergio: there should be something about incomplete information in the use case

<pfps> Jeremy: data integration use case is interesting to HP because it is simple

<pfps> Hassan: why is the use case related to incomplete information?

<pfps> Sergio: there is often some sort of incomplete information in data integration

<pfps> Sandro: This is a synthetic use case - does anyone have a real use case?

<pfps> Frank: yes, there should be lots - Fujitsu has done this

<pfps> Jeremy: let's do something that can be done fast, let's not try to perfect the use case now

<pfps> ?: production systems actually don't work on vocabulary mapping

<pfps> they work, as well, on data mapping

<pfps> Chris: Sandro will do the use case, Sergio to provide input

<pfps> ACTION on Sandro - effect Proposal 3 as modified

<pfps> Chris: discussion on Proposal 1

<pfps> ?: there is still something left in the area - something like distributed rule bases

<pfps> Harold: the use cases in 2.6 need homes

<pfps> Christian: FOAF use case ends up in 2.1

<pfps> Sandro: don't see anything valuable left in 2.6

<pfps> Chris: objections

<pfps> ACTION on Allan as editor of UCR - remove 2.6

<pfps> Hassan: status of Use Case titles in ToC

<pfps> Chris: proposal is to do this somehow

<pfps> Hassan: as long as it works out

<pfps> USE CASE 2.2: Ruleset integration for Medical Decision Support

<pfps> Chris: Status?

<pfps> Christian: change title to above, drop e-learning,

<pfps> concentrate on prescription part (and extend slightly)

<pfps> Chris: Objections

<pfps> ACTION on Chris: tell Leora to make changes

<JosDeRoo> ACTION: JeffPan propose extension to prescription example [recorded in http://www.w3.org/2006/02/28-rif-minutes.html#action03]

<JosDeRoo> discusssing UCR/Interchange of Human-oriented Business Rules

<JosDeRoo> merging with 2.6 is not an option

<JosDeRoo> ChrisW: use case is not close as it stands, needs catching better the intention

<JosDeRoo> .. propose to move it for next WD round?

<JosDeRoo> FrancoisB: have something about BR in first draft is crucial

<JosDeRoo> CSMA: propose to keeep title in first WD

<JosDeRoo> CSMA: alternatively Donald to come up with new text by monday?

<JosDeRoo> .. text can be removed

<JosDeRoo> RESOLVED no objection

<JosDeRoo> ACTION: JohnHall come up with new text by Monday [recorded in http://www.w3.org/2006/02/28-rif-minutes.html#action04]

<JosDeRoo> proposed title: tools for managing policies and practices in organizations

<JosDeRoo> (discussing a proposed title..)

<JosDeRoo> RESOLVED: new title "Managing inter-organizational business policies and practices"

<JosDeRoo> ACTION: JohnH to reflect that title in the new proposed text [recorded in http://www.w3.org/2006/02/28-rif-minutes.html#action05]

<JosDeRoo> DONE with UCR discussion!!! fireworks

<JosDeRoo> next topic: next F2F

<JosDeRoo> AxelP: proposal to colocate with next ESWC Montenegro June 11-14

<JosDeRoo> .. either 8-9 or 9-10 is proposed

<JosDeRoo> Igor: afterconference is not an option

<JosDeRoo> Igor: 8-9 was originally planned for RIF

<JosDeRoo> no objection to have next F2F Montenegro either June 8-9 or June 9-10

<JosDeRoo> RESOLVED: to have next F2F Montenegro either June 8-9 or June 9-10

<JosDeRoo> RESOLVED: to have WBS vote per organization

<JosDeRoo> ACTION: Sandro to organize such WBS vote [recorded in http://www.w3.org/2006/02/28-rif-minutes.html#action06]

<JosDeRoo> by next tcon have resolution on the date

<JosDeRoo> ACTION: AxelP to set up logistics F2F page [recorded in http://www.w3.org/2006/02/28-rif-minutes.html#action07]

<JosDeRoo> next topic: F2F4

<JosDeRoo> proposed to have host in NA

<JosDeRoo> .. around October

<JosDeRoo> CSMA: propose to submit proposals and resolve next F2F

<JosDeRoo> next topic: Roadmap

<JosDeRoo> presentation by HaroldB

<JosDeRoo> the presention is on the mailing list http://lists.w3.org/Archives/Public/public-rif-wg

<JosDeRoo> Frank: if you do assert then you have to be able to undo it as well

<JosDeRoo> (discussion about semantics of production rules)

<JosDeRoo> ACTION: Frank to make his point on the mailing list [recorded in http://www.w3.org/2006/02/28-rif-minutes.html#action08]

<JosDeRoo> Francois: need for complex 'actions' in production rules (besides complex events and complex conditions)

<JosDeRoo> Harold: can already do the syntactic extensions in Phase1

<JosDeRoo> .. to make clear what might be done semantically in Phase2

<GaryHallmark> +1 on including production rules earlier

<JosDeRoo> ACTION: Harold to explain technically the basis for interoperation between PR and Horn rules [recorded in http://www.w3.org/2006/02/28-rif-minutes.html#action09]

<JosDeRoo> MichaelK: Phase1 is one semantics i.e. FOL entailment (is in response to Uli's question)

<JosDeRoo> CSMA: will set up new page for requirements

<JosDeRoo> LUNCH

<MarkusK> http://www.w3.org/2005/rules/wg/wiki/UCR/Requirements

<JosDeRoo> discussing http://www.w3.org/2005/rules/wg/wiki/Draft_-_Requirements_on_RIF

<ChrisW> mike dean, are you here?

<mdean> yes

<MarkusK> Scribe: Markus (4pm -- 5pm)

<MarkusK> Sandro: Previous session of collecting requirements is not satisfying.

<MarkusK> ... there are many rule systems, but we need to identify common features

<MarkusK> ... there should be some sharing among similar systems

<MarkusK> ... e.g. sharing between several different production rulse systems

<MarkusK> ... Try to establish a framework

<MarkusK> csma: agrees. Identify features that can be reused by other systems.

<MarkusK> ... provide guidance of how to work with unsupported features.

Semantic Web compatibility

<MarkusK> Chris: Highly important to be compatible with existing Semweb standards

<MarkusK> ... different possible views:

<MarkusK> ... (1) OWL as a part of (the syntax of) RIF

<MarkusK> ... OWL does not get translated

<MarkusK> ... but is just loaded in RIF without change

<MarkusK> ... (2) OWL as a target (?) language

<josb> s/OWL as a target (?) language/OWL is a covered/source/target language

<MarkusK> Mala: What about other langues such as Common Logic?

<MarkusK> Chris: OWL ist a W3C standard, so compatibility is part of the charter.

<MarkusK> Harold: I would like to add a third point: (3) interoperation

<MarkusK> ... meaning that one refers to another (OWL) document which remains external

<josb> As I understood Harold, you include queries to OWL ontologies in the body of the rule

<MarkusK> Yes, I think so

<MarkusK> Sandro: There might be another aspect of (1): Transforming parts of OWL (e.g. DLP) into a rule syntax.

<MarkusK> Elaboration of point (2):

<MarkusK> * syntax/semantics of OWL in RIF

<MarkusK> * mapping of OWL "Rules"

<MarkusK> [literal transcription of board]

<MarkusK> Francois: +1 Harold (i.e. point (3) above)

<MarkusK> Hassan: It is not the concern of RIF to specify how to reason with encoded logical specifications. RIF provides the semantics and those who employ RIF must provide their operationalisation.

<MarkusK> ... It suffices to encode OWL semantics in rules.

<MarkusK> Francois: Syntax of course must be compatible or translatable -- but this is secondary.

<MarkusK> ... Reasoning is more important.

<MarkusK> ... Reasoning for rule languages is typically constructive.

<MarkusK> ... Which is simpler than reasoning in other formalisms (no excluded middle, refutation, ...)

<MarkusK> ... Implementing an OWL reasoner in RIF would not be good.

<MarkusK> Chris: We do not want to do this.

<MarkusK> Fraincois: Translating OWL into RIF does not make sense. (?)

<MarkusK> Harold: The expressiveness needed to translate OWL into RIF is not available in Phase 1.

<MarkusK> Chris: Any other ideas on Semantic Web compatibility as well?

<MarkusK> csma: Can this be applied to RDF as well?

<MarkusK> Harold: We probably mean OWL DL and this refers to a subset of RDF as well.

<MarkusK> Chris: I was talking about OWL in general.

<MarkusK> Chris: What are the options for RDF and RDFS?

<josb> Allen

<Darko> Frank

<csma> \me franck

<josb> Frank McCabe

<MarkusK> FrankMcCabe: The case of combining OWL with another rule language is important.

<MarkusK> .. e.g. one could combine OWL with Prolog

<MarkusK> ... and this combined language might have a different mapping to RIF

<MarkusK> Chris: Mapping OWL into Prolog is lossy already.

<MarkusK> FrankMcCabe: It might be possible to have non-lossy combinations.

<MarkusK> Chris: It seems that it would then be possible to use a uniform mapping.

<Darko> Hasan

<MarkusK> Francois: Rule languages often have no disjunctive reasoning.

<MarkusK> ... and translating C \sqsubseteq A\sqsup B might be problematic

<MarkusK> ... since a lossy translation "A <- C" is not useful

<MarkusK> Chris: Why should a rule language not have disjunctions in the head?

<MarkusK> Francois: Such rule languages do not seem useful, since they might be incompatible to existing languages.

<MarkusK> Sandro: Full FOL is in scope for Phase 2.

<MarkusK> csma: If there is an implemented rule language that allows disjunction in the head, then we can decide on whether we want to cover these or not. We do not question the language.

<MarkusK> Discussion on whether OWL syntax should be part of RIF.

<MarkusK> Chris: This may conflict with the requirement of having a nice syntax.

<MarkusK> Deepa: If OWL/XML is part of RIF, does every RIF-conformant processor have to support OWL?

<MarkusK> Chris: No. This is not required, since hardly any system can be expected to support all languages captured by RIF.

<MarkusK> Sandro: The label "sublanguage" [used by Chris earlier] was replaced early on by the idea of "modules" in the charter.

<MarkusK> ... So one can have systems conforming to some RIF module.

<csma> \me jos de roo

<MarkusK> Sergio: Not every OWL-DL document is an RDF document. The syntactic way in which OWL-DL is defined is not very simple. Including it directly in RIF might cause unpleasant definitions of syntax.

<MarkusK> Chris: The basic question is: should there be a "special treatment" for OWL, or is it just treated as any other language we support?

<MarkusK> Sandro: The options (2) (a) [OWL semantics in RIF, new syntax] and (b) [OWL translated into rules, that are more expressive than Horn] are available for any language.

<MarkusK> Francois: Theory Reasoning/Theory Resolution might also be a general option.

<MarkusK> Hassan: This formalism was made obsolete by Constraint Programming.

<pfps> Christine

<MarkusK> Christine: What does "greater-than-Horn expressivity" mean in point (2)(b)?

<MarkusK> ... will we end up with a super-set approach when supporting many languages in this way?

<MarkusK> Chris: OWL has a special status that other languages *may* not have.

<MarkusK> Francois: We are at the crossing point of this WG!

<MarkusK> ... Two choices:

<MarkusK> ... (i) RIF is a language which has disjunctive rules (disjunction in heads). This has advantages (great expressiveness), but also disadvantages: you may not find efficient reasoners.

<MarkusK> Sandro: There might be more reasoners in the future.

<MarkusK> ... It must not be our concern at the moment.

<MarkusK> Francois: It is still dangerous. we might come up with things that are not supported by any usable reasoner. It might not be realizable in practical systems.

<MarkusK> ... this is rather a reasearch topic.

<MarkusK> Chris: We do not require that there are reasoners for RIF.

<MarkusK> Francois: Second approach:

<MarkusK> ... (ii) We can look around at best practices in Computer Science.

<MarkusK> .. there are currently three kinds of rules.

<MarkusK> ... Database-like rules, Integrity constrainst, and

<aharth> reactive rules

<MarkusK> ... So one should look at existing tools to find what is useful in practice.

<MarkusK> Chris: Let us turn our findings into something concrete.

<MarkusK> ACTION: Francois writes down the details of his proposal. [recorded in http://www.w3.org/2006/02/28-rif-minutes.html#action10]

<MarkusK> Chris: Other approaches to Semweb compatibility?

<MarkusK> pfps: You could relax the requirement of having "OWL/XML" as a syntactical embedding of OWL into RIF.

<MarkusK> ... so why not admit another syntax, such as OWL abstract syntax.

<MarkusK> Chris: This is not an option since OWL/XML is the only syntax for OWL Full.

<MarkusK> == Vote ==

<MarkusK> Who thinks (1) is the way to support OWL?

<MarkusK> -> 1 person

<MarkusK> Francois: In FOL you often have a placeholder for terms, whereas in OWL you leave terms implicit.

<MarkusK> ... I like both. So maybe we can keep both options?

<MarkusK> ... you can choose either way, or try to combine both.

<MarkusK> ... Description Logic has another cultural style than FOL.

<MarkusK> ... There might be problems of combining the two.

<MarkusK> Michael: Clarification: (1) was embedding without translation, and (2) (a) includes a new syntax for OWL?

<MarkusK> cmsa: Yes.

<MarkusK> Chris: So there a two styles, one based on modal logic that shows not terms, the other in FOL with explicit terms.

<MarkusK> ... but the modal syntax can be mapped to FOL.

<MarkusK> ... So it probably not that hard to map one style to the other.

<MarkusK> Chris provides more clarification concerning the different between (2) (a) and (b).

<MarkusK> Who thinks that (2) (a) is the way for RIF to be compatible with OWL?

<MarkusK> -> 7 people

<MarkusK> Who thinks that (2) (b) is the way for RIF to be compatible with OWL?

<MarkusK> -> 15 people

<MarkusK> Who thinks that (3) is the way for RIF to be compatible with OWL?

<MarkusK> -> 18 people

<MarkusK> csma: In (2) (b) OWL implementers seem to have to do the work, while in (a) we seem to have to do it. That is why I prefer (b).

<MarkusK> MichaelKifer: Nobody has to do the work.

<MarkusK> Sergio: The works by Eiter and Rosati are examples for (3).

<MarkusK> ... both are possible.

<MarkusK> Igor: I had a different undertanding of (3).

<MarkusK> ... Wasn't (3) more like SPARQL?

<MarkusK> Chris: No, rather not.

<MarkusK> JosB: If you have a FOL sublanguage of RIF, then (2) is possible. But in a Logic Programming setting, you have to restrict to (3).

<MarkusK> Chris: Any other comments? [none]

<MarkusK> ... Another aspect of the compatibility discussion are URIs.

<MarkusK> pfps: We should use IRIs

<MarkusK> Sandro: I think we can still say "URIs"

<MarkusK> Chris: So to what part are URIs part of RIF?

<MarkusK> Chris: Should all our symbols be URIs?

<MarkusK> Francois: The URI issue seems to be important for reasoning on the web.

<MarkusK> ... Another aspect is privacy of data. So can we restrict RIF rules to some part of the Web?

<MarkusK> ... Something similar is possible in RDF. There one can include statements.

<MarkusK> [General disagreement: there is no import/include in RDF]

<MarkusK> Sandro: URIs are used both as names and as locations.

<MarkusK> ... RDF never uses them as addresses. We have to be careful to distinguish this.

<MarkusK> Chris: To the commercial rule vendors:

<MarkusK> ... would it be a limitation to have to use URIs for every symbol in RIF?

<MarkusK> FrankMcCabe This is just a mapping issue. No major problem.

<MarkusK> Francois: The possibility of hiding is very important. It might be useful for us.

<MarkusK> ... (hiding as in Software Engineering)

<MarkusK> Sandro: We would need a serious reason for supporting this.

<MarkusK> ... It seems to be complicated.

<josb> +1

<MarkusK> Frank: The descision that some identifiers are the same can be complicated in some logical systems.

<MarkusK> ... So requiring that every URI is a different thing affects reasoning.

<MarkusK> Chris: E.g. in OWL this is not assumed.

<MarkusK> Scribe job passed on to MichaelSintek

<msintek> Harold: talk more about RDF

<MarkusK> The above "This is just a mapping issue. No major problem." was said by Paul Vincent.

<msintek> Harold: bnodes, see wiki

<josb> ScribeNick msintek

<tlr> ScribeNick: msintek

Harold: binary/ternary relations, ...
... mappings; correspondence of RDFS and RIF

Chris: any burning issues?

csma: major achievements: use cases ...
... design goals ...
... moving toward technical spec
... thanks the scribes :-)
... minutes as fast as possible
... reminds us of schedule
... telecon in two weeks: uc
... lots of actions
... closing statement: thanks everybody
... see you next time

- DRAFT -

RIF F2F2, day 2

28 Feb 2006

Attendees

Contents

Semantic Web compatibility

Summary of Action Items

Scribe.perl diagnostic output