SV_MEETING_TITLE -- 4 Apr 2006

I guess that I can do it

<igor> this is me

<igor> no

<igor> sorry

Next meeting same time same place

Christian moves to accept minutes

of last meeting

<PaulaP> +1

<Hassan> +1

<josb> scribenick: IanH

Seconded by PaulaP and Hassan

Resolved to accept minutes

<PaulaP> sorry for the delay

Scribes asked to distribute munutes asap and if pos within 48 hrs

Any ammendments to agenda?

<sandro> ACTION: [PENDING] chair to put design for extensibility and discussion of proposals on agenda for next telecon [recorded in http://www.w3.org/2006/03/28-rif-minutes.html#action14]

Harold wants to add item on new members

<sandro> Alex Kozlenkov

<sandro> from Betfair

<scribe> New member Alex Kozlenkov introduced himself

<scribe> ACTION: Public membership list needs to be extended to include Alex [recorded in http://www.w3.org/2006/04/04-rif-minutes.html#action02]

Christian: Alex can introduce his background via mailing list

<sandro> http://www.w3.org/2005/rules/wg_participants.html

<AxelPolleres> Propose Alex checks intro mails in the archive and sends a similar one.

<sandro> http://www.w3.org/2000/09/dbwg/details?group=38457

<sandro> (since it's already done)

F2F meetings

2 proposals for f2f4 (one with 3 options)

deadline for decisions May 3rd/16th

will therefore make decision by May 2nd mtg

let us know by april 18 if you have any proposal for f2f4

will review on April 18 and set up straw poll closing May 1st

Decision will be taken at May 2nd mtg

Process agreed by popular consent

Next topic: action for Alan to put Mitre proposal on Wiki - DONE

Next topic: liasons - any news?

<JosDeRoo> zakim. unmute me

Any SPARQL participant who is more often on teleconf? Jos?

<sandro> ACTION: [DONE] Allen to put MITRE proposal on f2f4 wiki page [recorded in http://www.w3.org/2006/03/28-rif-minutes.html#action01]

JosDeRoo: SPARQL proceeding apart from small objection from Oracle

Discussion on using UTC for telecon times

<scribe> ACTION: Christian will investigate [recorded in http://www.w3.org/2006/04/04-rif-minutes.html#action04]

Christian: XQuery/XPath news? Common Logic?

Who was that?

Evan: no important news.

<EvanWallace> New version of ODM: http://www.omg.org/cgi-bin/doc?ad/2006-01-01

Donald: no important news on SBVR

Evan: new draft on ODM

Christian: any other news from any other group?

No news.

Next Topic: UCR

Christian to modify template: DONE

Also made slight modificaton to structure

<sandro> ACTION: [DONE] csma to modify design constraint template to take into acccount what was said today [recorded in http://www.w3.org/2006/03/28-rif-minutes.html#action12]

<sandro> ACTION: [PENDING] Evan to publicize to ODM [recorded in http://www.w3.org/2006/03/28-rif-minutes.html#action09]

<johnhall> did it, said so last week

<sandro> ACTION: [DONE] John Hall to publicise to BR community [recorded in http://www.w3.org/2006/03/28-rif-minutes.html#action04]

<sandro> ACTION: [DONE] JosB to publicise to SPARQL [recorded in http://www.w3.org/2006/03/28-rif-minutes.html#action05]

<sandro> ACTION: [DONE] Sandro to set up a wiki page to record dissemination actions [recorded in http://www.w3.org/2006/03/28-rif-minutes.html#action10]

Christian: any comment on dissemination?

Sandro: only one comment on comments list

<sandro> http://www.w3.org/2005/rules/wg/wiki/Outreach

Christian: not yet evaluated so will discuss next week or via email

<sandro> comments archive: http://lists.w3.org/Archives/Public/public-rif-comments/

Next Topic: discussion of specific constraints

Christian: is there a way to visualise structure if we draw dependencies between design constraints? Any visualisation tool for this?

Frank: I use Omnigraffle(?) a Mac tool

<SaidTabet> Frank: Can you point us to it? URL?

<scribe> ACTION: Frank will produce an initial diagram with existing constraints [recorded in http://www.w3.org/2006/04/04-rif-minutes.html#action10]

Frank: it has its own format, but will generate visio, jpeg, bmp etc etc
... but probably need Omnigraffle to modify (maybe via visio)

Christian: how do we decide which constraints to discuss, particularly when list grows
... proposes only to discuss constraints that have been seconded

Any objections?

Resolved by popular apathy/aclaim

<sandro> http://www.w3.org/2005/rules/wg/wiki/Design_Constraints

Christian: Dave Reynolds proposed strawman goals hierarchy; I believe that it is useful and will help structure constraints
... wonders why almost no goals belong to existing list?

<sandro> dave's email: http://www.w3.org/mid/44326F35.9030804@hplb.hpl.hp.com

Dave: was mainly trying to get structure/level right rather than map existing list.
... believes that some of existing goals *are* in the list, e.g., production rules, semantics, compatibility
... list focuses on high level goals
... wanted to avoid swamping with detail

<sandro> +1 coherent presentaiton in e-mail vs putting this particular hierarchy on the wiki as is

Christian: encourages people to react to email and comment and add to requirements if believed to be important
... Is clear and precise semantics realy a requirement.

<sandro> FrankMcCabe

<JosDeRoo> Sandro: I was wrong with fixed utc start times for rdfcore and webont; is only dawg that has fixed utc start time and I generalized; sorry

Frank: Probably not if we can't measure it

<sandro> Good, Jos -- I didn't think I'd misunderstood common practice here so badly.

Frank: Wording could be improved - e.g., success will depend on having clear semantics of a RIF rule set
... should be sharpened up and made a requirement

Dave: what does it mean for a proposal to be seconded?

Christian: agrees that dependency on another constraint constitutes support for that constraint

Leora: constraints all seem straightforward and uncontroversial.

<sandro> http://www.w3.org/2005/rules/wg/wiki/RIF_Core_must_cover_pure_Prolog

Christian: may not be true w.r.t. phasing

<igor> +1

<JosDeRoo> +1

<AxelPolleres> Do you mean full ANSI prolog?

Christian: second for pure prolog use case?

Seconded by Igor

<sandro> Extended RIF must cover FOL

<MoZ> +1

<EvanWallace> +1

<LeoraMorgenstern> Ian, see http://www.w3.org/2005/rules/wg/wiki/Design_Constraints

2nd constraint: must cover 1st order logic

Seconded by NoZ and Evan

3rd constaint: must be sound

Harold: not sure this makes sense - all logics are sound

Allen: soundness may be a requirement on systems rather than RIF

<josb> How can you reason with something which is unknown?

Christian: requirement should be rephrased to clearly be a requirement on RIF and not on systems

Allen: agreed; also, do we want requirements/constraints on systems?

<LeoraMorgenstern> Ian, in your scribing, you say 2nd constraint mentions 2nd order logic; actually, it mentions first order logic!

<LeoraMorgenstern> (confusion of seconding the constraint and the order of logic ;)

CORRECTION: second constraint refers to 1st order logic (thaks)

Christian: nobody has yet seconded the soundness constraint; should be clarified and discussed on email before being discussed on teleconf

Sandro: disagrees with procedure; teleconf is good place to discuss focus of constraint (RIF -v- systems)

Christian: agrees in general, but w.r.t. specific constraint 2 clarifications needed: what does "sound" mean; is it a RIF or system constraint

<scribe> ACTION: sandro to clarify meaning of sound and what is the requirement on RIF [recorded in http://www.w3.org/2006/04/04-rif-minutes.html#action11]

Allen: requirements often reflect proposed/imagined applications

<sandro> (I'll probably rename it: RIF Core must allow sound reasoning with unknown dialects

<sandro> )

Allen: but this may lead to revealing constraints/requirement on RIF

<AxelPolleres> -1 (sorry very late) to RIF_Core_must_cover_pure_Prolog: Prolog is not declarative, I don't like that in RIF core

Allen: So may be useful to talk about applications w.r.t. critical success factors and/or goals

Harold: Asks clarrification question that I didn't understand w.r.t. "soundness"

<GaryHallmark> -1 to RIF_Core_must_cover_pure_Prolog -- this implies that every prolog ruleset must be able to be translated to every rule dialect we support. I'm not sure that's possible with production rules

Sandro: rephrase requirement 3 as RIF core must allow sound reasoning with mixed dialects

<Harold> Sandro, Re candidate DC "Sound reasoning with unknown dialects" from the RIF perspective we could say "Permit heterogeneous rules in a single ruleset".

<JosDeRoo> use case?

<josb> +1

<Allen> skipped one

<DaveReynolds> You skipped one, the next one was The RIF Core must be able to accept RDF triples as data

<scribe> ACTION: to clarify whether sound reasoning constraint with unknown dialects is a requirement or a critical success factor [recorded in http://www.w3.org/2006/04/04-rif-minutes.html#action12]

<AxelPolleres> +q

<JosDeRoo> +1

<sandro> You want "q+" Alex.

<josb> yes

<MichaelKifer> regarding RIF_Core_must_cover_pure_Prolog: Pure prolog = Horn. I understand that we are committed to that

Constraint: RIF must be able to accept RDF as data

<sandro> You want "q+" Axel

<JosDeRoo> yes

<DaveReynolds> +1 to The RIF Core must be able to accept RDF triples as data

<AxelPolleres> +1 to RDF

Seconded DaveReynolds

<josb> I will second this as well

And seconded by Axel

Sandro/Christian: seconding should be via web page

Christian: you have to add your name on the wiki page for that constraint

<AxelPolleres> Will there also be the possiblity to *object* on the Wiki page?

<sandro> http://www.w3.org/2005/rules/wg/wiki/RIF_Core_must_cover_pure_Prolog

Christian: open discussion on pure prolog

<AxelPolleres> I think jos will clarify this...

<Uli> can we refine constraints? If yes, what happens to the secondings?

Christian: Axel didn't support?

Josb: problem is that prolog doesn't have declarative semantics; would require procedural semantics for RIF

<Allen> that depends what you mean by "cover"

Christian: too weak?

<MarkusK> An interesting link considering declarativity in Prolog: http://www.cs.sunysb.edu/~warren/xsbbook/node2.html

<AxelPolleres> IanH: I didn't support prolog in the sense of jos

Josb: no, too strong; would requirem procedural semantics.

<Harold> MichealK, I agree in the sense that Pure Prolog 'realizes' Horn, which is the language of the Charter's Phase 1.

<FrankMcCabe> I think that jos is misunderstanding the situation

Josb: semantics only defined in terms of algorithm; if this is a req. then we would need this kind of algoritm to define semantics for RIF core

Sandro: e.g. requiring SLD resolution or some such

Gary: Agree, it is too big, and this would mean that every prolog rule set would need to be translated into production rules.
... believes that this would lead to problems

<AxelPolleres> I was just about to mention the example that "pure" Prolog doesn't terminate on queries that can be fulfiled and depends on the rule order, so a ruleset, is actually not a set there.

<csma> ack

<MoZ> +1 to define pure prolog

Peter: two issues. 1) what is defn of pure prolog? idealisation of prolog without negation etc. This is effectively horn clauses which is better. 2) this is only semi decidable, which is bad for core

Sandro: believes that by pure prolog he really meant Horn
... thinks that procedural semantics may be needed in real world examples.
... needs clarification; we are getting there.

<Uli> ...and perhaps even function-free Horn?!

<igor> pure Prolog = logic program (Horn) with defined order of goals and clauses

Harold: believes that RP realises prolog; can imagine lex. ordering not taken into account; needs some procedural stuff

<MoZ> i imagine with have to settle an action to define the subset of Prolog we cover in RIF language

<josb> +1 to start we Datalog

<josb> Datalog also has a declarative semantics

Harold: decided in charter to specify Datalog

<AxelPolleres> +1 to igor's definition and thus -1 to the requirement! :-)

Sandro: pure prolog covers Datalog

<AxelPolleres> Datalog is declarative, i.e. rule order does not matter!

Harold: was discussed 1 year ago and we decided to start with Horn and not just datalog

Christian: reason was to have functions in phase 1

Harold: this makes it undecidable

Sandro: maybe we should separate requirements into Datalog and pure prolog

Harold: what about charter?

Christian: again, because we wanted functions

Frank: Harlold said it all.
... Most people believe that PP means Horn with no procedural semantics
... has been a religious argument for many years

<josb> Let's use an unambiguous terminology!

Alex: In RIF core will there be any way to mark rules as using a different semantics, or will there be a uniform semantics for all rules?

<MarkusK> What does it mean to "implement the RIF core", for a semi-decidable core?

Christian: every implementation must implement at least RIF core, but not all implementations will be able to implement multiple semantics - big burden on implementors

Alex: but what about provision for plug-in semantics?

<Uli> ...but how will this be an implementation for their combination? We would need to define what their combination means?!

<FrankMcCabe> Semi-decidable = soundness and completeness and decidable equality?

Christian: believes this is for extended RIF, but could be discussed

<MarkusK> +1 to Uli

Michael: believes PP is equivalent to Horn rules

<LeoraMorgenstern> am I the only one hearing french on the line?

Michael: is there a requirement for decidability of Core?
... what is wrong with core being undecidable?

<Uli> Semi-decidable: "yes" will be answered correctly, "no" might not terminate

<sandro> no, LeoraMorgenstern, me too

<MoZ> sorry

Michael: at f2f we proposed semantic tagging of rule sets for extensibility

Michael/Alex: this is needed for extensibility

<MarkusK> Semi-decidable: all "yes" are recursively enumerable

<FrankMcCabe> there is a completeness proof for SLD logic

Machael: only question is whethere core needs to be decidable; what about production rules? not compatible with Horn

Allen: Agrees with Michael; no need to be decidable.
... notion of representation without being able to reason

Christian: what do we do with non-requirements?

Frank: good idea to have non-requirements

Sandro: we can just decide this; can be written up on wiki page

Christian: this may apply to decidability (it may be a non-requirement)
... wants write up as to why RIF core should *not* cover pure prolog

Sandro: would like to separate this from decidability (non-)requirement

<GaryHallmark> does anyone know how to characterize the intersection of production rules and pure prolog?

Sandro: question as to definition/semantics also needs to be clarified

<AxelPolleres> If it was about specifying the criticicm, I can do that. In the document?

<scribe> ACTION: MickaelK to extend page on pure prolog and give a precise definition (according to standard publications) [recorded in http://www.w3.org/2006/04/04-rif-minutes.html#action13]

<sandro> page is http://www.w3.org/2005/rules/wg/wiki/Pure_Prolog

Christian: volunteer to clarify (non)decidability?

Sandro: proposes that decidability is not a requirement

<AxelPolleres> We should of course define the decidable fragments and give them a name. Or no?

Uli: could we be clearer about what kind of reasoning problems we are talking about?

Christian: agrees; this needs to be discussed and clarified

<sandro> +1 define somewhere the decidable fragments, w.r.t. particular reasoning tasks, like query answering

Axel: even if we decide against decidability, would be useful to identify known decidable fragments (w.r.t. particularl reasoning tasks)

<Harold> Uli, right: issue is decidability of conjunctive queries

<Uli> +1

<Harold> Axel, yes: standard sublanguage Datalog is a good idea.

<pfps> A counter to Michael's claim that PP is always Horn (only) can be found in http://www.cs.nyu.edu/courses/spring02/G22.2560-001/horn.html

Christian: would like someone to propose decidability (non-)requirement so we can discuss something more precisely defined

Harold: Charter says Horn, and this is undecidable

<josb> +1 to pfps; we should be unambiguous in the terminology

<AxelPolleres> +1 to Hassan

<josb> +1 to Hassan

<MarkusK> +1 to avoid the term Horn

<PaulaP> +1 to Hassan's comment

<Uli> +1

<MarkusK> s /Horn/Prolog/

<sandro> Moz, I muted you because you were being noisy

<AxelPolleres> MArkusK, you meant Prolog?

Hassan: hears a lot of confusion; shouldn't use prolog, should use Horn; not clear what pure prolog means - should just cast out this term

<MarkusK> Axel: yes

Hassan: any interesting rule language will be undecidable, so its a non-requirement

<MichaelKifer> +1 to Hasan's proposal (especially since this obviates my action item :-)

Hassan: need rule annotations that can specify semantics (which may be declarative or procedural)

<scribe> ACTION: Christian to send email to propose resolution that decidability is a non-requirement and gather relevant arguments before next telecon [recorded in http://www.w3.org/2006/04/04-rif-minutes.html#action14]

<sandro> Extended RIF must cover jrules

Sandro: Imagines each participant requires that extended RIF covers their rule language of choice
... thought that pure prolog was, however, small enough to be covered in core, but is a real language that is widely used and well understood

<josb> -1 to Sandro

Harlod: agrees; can extend such an implementation in many directions.

<Uli> ...but prolog doesn't have a declarative semantics!

<sandro> +1 extend 10 minutes

<MoZ> +1

<DaveReynolds> -1 continue by email

<josb> -1: continue by email

<sandro> resolved 5 minutes

<AxelPolleres> If we agree to Sandro then we implicitly should agree to drop the term rule-"set" odr no?

At *most* 10 minutes for me!

<josb> Few people agree with Sandro!

<josb> +1 to Hassan again

<Harold> ...Pure Prolog (with annotation: no textual order, breadth-first, occurs check) has declarative semantics of Herbrand models.

Hassan: pure prolog is not a language; it is a set of rules and you need to say how they will be used; let's not be ambiguous.

<josb> ++++++++1

<Uli> Harold, I thought "Pure prolog" is depth first, and order-dependent?!

MichaelK: Peter points our varioations on meaning of pure prolog. May mean terminating prolog rules. Michael agrees; suggests we only talk about Horn. He will clarify on wiki.

<AxelPolleres> I thought the same as uli.

GaryH: Agrees with Sandro about favourite language; core must be common to everyones favourites; is "pure prolog"/Horn small enough to be in this intersection?

<pfps> The pointer I gave does define Pure Prolog as L-R evaluation, and other deviations from "Horn" semantics.

<FrankMcCabe> we must decide on declarative vs procedural semantics

Sandro: will be interested in outcome of discussion.

Christian: aob?

<PaulaP> bye

Christian: hearing none, meeting is closed.\

<Deborah_Nichols> bye

<Uli> bye

<JeffP> bye

<Allen> bye

<SaidTabet> bye'

<igor> bye

<MoZ> bye

<MalaMehrotra> bye

<Harold> Uli, in John Lloyd's "Foundations of Logic Programming" depth first is labeled as a search strategy that leads to implementation incompleteness.

<csma> make minutes

<AxelPolleres> because, he seems to neglect the difference...

<AxelPolleres> ooops

- DRAFT -

SV_MEETING_TITLE

4 Apr 2006

Attendees

Contents

F2F meetings

Summary of Action Items

Scribe.perl diagnostic output