W3C

- DRAFT -

RIF Telecon 31 July 07

31 Jul 2007

Agenda

See also: IRC log

Attendees

Present
+39.047.101.aaaa, ChrisW, Sandro, Hassan_Ait-Kaci, Dave_Reynolds, StellaMitchell, LeoraMorgenstern, josb, Harold, Gary_Hallmark, Allen_Ginsberg, AxelPolleres, IgorMozetic, luis_polo, MichaelKifer
Regrets
Fran´┐ŻoisBry, PaulaLaviniaPatranjan, JeffPan, MichaelSintek
Chair
Chris Welty
Scribe
Hassan

Contents


 

 

<ChrisW> list agenda

<ChrisW> scribe: Hassan

<ChrisW> scribenick: Hassan

Admin

Objections to last week minutes?

<Harold> Did we approve the f2f6 minutes?

None

Minutes accepted.

Next telecon August 7 Poll

<LeoraMorgenstern> I can be here.

<sandro> +1 plan to be here Aug 7

<josb> +1

<DaveReynolds> -1

<Harold> +1

+1 plan to be here (hak)

<GaryHallmark> -1

August 14?

<LeoraMorgenstern> +1 to the 14th

<sandro> +1 plan to be here Aug 14

<josb> +1

<DaveReynolds> +1

+1

<Harold> +1

<StellaMitchell> +1

Auguts 21?

<sandro> +1 aug 21

<josb> +1

+1

<Harold> +1

<DaveReynolds> +1

<LeoraMorgenstern> +1

<StellaMitchell> +1

<GaryHallmark> +1 for 21/8

Aug 28?

<LeoraMorgenstern> +1

<sandro> +1 Aug 28

+1

<Harold> +1

<DaveReynolds> + 0.5

<josb> +1

<GaryHallmark> +1

<AxelPolleres> 14th -1, 21 +1, 28 +1

Chairs will discuss the schdule for Aug telecons and will get back to the rest

Action reviews...

Action 332 ?

Dave Reylnolds - has to do with classification.

Action 332 complete

Action 329 on ChrisW complete

Action 324 - continued

Agenda amendments?

F2F

None heard

<josb> must be CTIC

F2F meetings - open actions

See wiki page

Questions?

<AxelPolleres> is the date now fix?

W3C tech plenary week of Nov 6 - give opportunity to RIF to attend?

Where- Boston.

We should decide by Sept ... ?

Sandro is confused about the dates ...

<sandro> I am not at all confused about the dates.

ChrisW: What value is it to have two meetings a month apart?

Sandro: need time to work on what to show in November
... we need a quorum for decisions

<Harold> F2F7 was pushed further and further into late Aug., then Sept.

<sandro> It's not about quorum -- it's about not excluding people.

ChrisW: the F2F will assign actions - decisions will be made by telecon

<AxelPolleres> 2 subsequent meetings in US, difficult -1

ChrisW: anyone having things to say on this issue?

<Harold> Can a F2F be replaced with a highbandwidth2highbandwidth video conference?

<DaveReynolds> If it would be useful but I couldn't guarantee to be there

Hearing none ...

Liaison

<sandro> http://www.w3.org/2007/11/TPAC/

ChrisW: moving to Liaisons

Report on liaisons?

Nothing on liaison

BLD - Sorts

Action review

331 on Chris - on mappings?

<scribe> continued

Action 330 on Dave - done

Action 330 kept open until discussed?

Action 300 is kept open for discussion

Action 328 continued (on Sandro)

Action 328 continued (on Sandro) on XML Schema Generation new date

Action 322 on Dave is completed

Action 300-301 done

ChrisW - no more action to review

Having the sort discussion and the frame classification discussion or the RDF?

BLD - Frames/classification

Michael being asked his reaction to the mail discussion on the subject

Michael summing up: we need a minimalistic model of class hierarchies for RIF

This will allow users using RIF for exchange to "hook" their models onto that of RIF's

This will preclude n^2 model exchange models

ChrisW: some existing for the SW - why reinvnet a new one? or why complicate the mapping from/to RDF

Any other?

Jos: RDF is close - but OWL is different (has extensional semantics ...)
... this proposal does not unify these views

MK: RDFS is a superset of that of RIF - so easy to extend

<ChrisW> ^existentials^extensional semantics^

MK: RIF class model is minimal in that it is easy to map onto all others

Jos: need additional axioms...

MK: no need

Jos: does this not imply a restriction?

MK: no ... RDF maps into RIF very naturally without problem

Dave: issues on scoping

MK: Scope for inference - what rules are visible

<Harold> Chris, since the Interop WS we have discussed a scope/module construct for rules, especially for scoped naf.

MK: classes and modules are linked
... no modules yet - but we are looking ahead
... want to avoid unintended interactions between imported KBs

ChrisW: let us focus on subclassing not on modules

<Harold> A scope/module is a 'local' ruleset, possibly named (as in RDF's named graphs).

<MichaelKifer> yes, the notion of modules/scopes is a generalization of rdf's named graphs

Dave: there are differences that may conflict among different models of inheritance
... second point - how much "subclass semantics" is there in RIF?
... I thought of RIF as a blank substrate but it does not look like it is any more

MK: there is still work to do on the specs

ChrisW: try to find the most general axioms of subclassing
... issue - why invent a new class model for semantic web?

MK: none of the existing model is minimalistic

We need a model that can be a core on most

MK: minimal in the number of axioms

ChrisW: what about OWL-DL?

MK: It has additional axioms (such as extensionality, ...)
... the minimal axioms need to be made explicit

Gary has a question - how do we express having no subclass?

MK: this is NAF
... this a meta statement that needs something like NAF

Dave - Gary: how can you have this in the head of a rule?

Dave - cannot change the type hierarchy dynamically

Dave - unless having meta model that allows you to

MK: maybe not an issue - e.g., interfacing with Java

<GaryHallmark> for Java data model, no membership or subclass relations in rule heads

Dave: need to assume the "right" set of axioms needed for the task

<GaryHallmark> XML schema data model has similar issues to Java

Sandro: it sounds like "minimal" means underspecified or ambiguous - if so that may be an interop problem

<josb> minimal subset of RDFS, which probably gives us all we need: http://www.eswc2007.org/pdf/eswc07-munoz.pdf

Sandro: expectations may be thwarted

MK: just transitivity of subclassing is enough

<sandro> MK: B subClassof A

<josb> structure?

MK: we do need inheritance of "structure" in RIF because we have no "structure" in RIF

<sandro> Okay -- I think I understand what Michael is saying now, about how RDF doesn't have inheritance. Just inheritance of the type -- not of the properties, etc.

ChrisW: what do you mean by structure?

MK: RDF has structure (slots ...) which is inherited
... Sorry - RDF has no inheritance

<sandro> so.... are # and ## proposed to be exactly the same as rdf:type and rdfs:subClassOf ?

<AllenGinsberg> +q

<sandro> Chris: similar to RDF, except that relation itself is not in the domain.

<AxelPolleres> Summary: transitivity of subclassing and of classmembership over the class hierarchy is the minimal semantics, yes? No more need to debate (for RIF) about additional axioms/inheritance notions.

ChrisW: minimal semantics of RIF subclass has not notion of structural inheritance

MK: yes

<AxelPolleres> reflexive, you mean?

<sandro> jos: If you restrict the syntax of RDF, you get something the same, except that the subclassof is reflexive. In RDFS if you know A is a class, then you know A is a subclass of itself.

Jos: RDFS subclassing is reflexive

<sandro> MK: That bit of RDF has been criticized and has no justification.

<sandro> (I'm comfortable with Jos's answer)

MK: adopting RDF as a model will entail adopting relations and we do not want to do that

Jos: why not use the core that already exists

MK: because we need to differentiate the RIF from any client models

<AxelPolleres> That would bring us back to the extensibility discussion, ie. whether "restricting" a dialect is covered by "extensibility".... or no?

<AxelPolleres> ... was referring to the comment that Gary's language would need restrictions to fit into BLD.

MK: why standardize things that many do not use and "impose" it on all

ChrisW: Are logic languages conform to this approach - can they be mapped to the minimal model?

MK: They will be translated to something else - not RDF
... they will have to map to different relations

ChrisW: is there really a problem?
... there are 2 aspects of RDFS subclass model missing from the RIF subclass model
... (1) syntactic reflection and (2) subclassing is reflexive

<josb> no

<DaveReynolds> It's in OWL

Allen: what about antisymmetry?

Jos - not in RDFS.

<AxelPolleres> http://en.wikipedia.org/wiki/Antisymmetric_relation

ChrisW - it is so in OWL-DL.

MK: these differences justify taking a miminal model of subclassing

ChrisW: those arguments in favor OO models in too early - keep in mind logic languages

<sandro> Chris: These differences between the proposed RIF Subclass and RDF Subclass .... RIF having a new, more minimal one, ... does this make sense?

Jos thinks the difference between RDFS/RIF subclass model is is not sufficient justification for introducing a classification vocabulary in RIF

Dave - having no issue except that could be even more minimal

<sandro> DaveReynolds: The question is what this stuff buys us -- why should it be in BLD at all? Just define the relation yourself, if you need it, when writing/translating the rule.

ChrisW: we seem to understand this issue better - move on ...
... any comment on this thread?

Igor: it would be good to have a resolution

ChrisW: good point -

Gary: I like the current proposal - except that we could fix it a bit for what we need

<josb> Gary, you might as well use RDFS

<josb> in both cases you need the same kind of restrictions

<sandro> Straw poll on each question....

ChrisW: (1) do we have a subclass op or (2) not

<sandro> +1 have subclass operator (instead of each one define their own mapping)

<LeoraMorgenstern> +1

<Harold> +1

<IgorMozetic> +1 for subclass

+1

<DaveReynolds> -1

<AllenGinsberg> +1 (only if extensible)

<GaryHallmark> josb: I'd need to see some kind of a proposal before I could agree

<luis_polo> +1

<sandro> MK say +1, not on IRC

<GaryHallmark> +1

<AxelPolleres> +1 subclass (as syntactic sugar foe a ruleset emulating the semantics of RDFS, basically)

Jos: does not understand the question

<josb> Gary, if you could send me a link to a use case, I can try to show how to use RDFS in this case

Axel: RDFS semantics can be emulated through ruleset?

<josb> In any Datalog-like language

MK: yes - it can be

<sandro> Axel: since RDFS can be expressed (emulated) as a BLD ruleset, then we can see this proposal for a syntactic construct of subclass, then this gives you an automatic inclusion of part of the RDFS ruleset.

<sandro> MK: I understand Axel to be supporting my proposal.

MK: Axel's point is in agreement with my proposal

<josb> My problem is that I need the additional constructs

MK and Axel argue a technical point ...

<josb> right

Nobody undertands ... :-)

<sandro> Axel -- what Sandro said is what I meant.

<AxelPolleres> yes, with emphasis on *part of* the RDFS ruleset.

<sandro> Axel: What MK wants is the same subset of RDFS that I have in mind.

<josb> reflexivity is the difference

MK and Axel: argue subclass operator in RIF and RDFS

<josb> yes

<AxelPolleres> probably yes.

ChrisW: will summaruze in email

Meeting over

<sandro> AxelPolleres, can you write the rules you have in mind, and send them in e-mail?

<DaveReynolds> bye

Summary of Action Items

[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.128 (CVS log)
$Date: 2007/07/31 16:36:59 $

Scribe.perl diagnostic output

[Delete this section before finalizing the minutes.]
This is scribe.perl Revision: 1.128  of Date: 2007/02/23 21:38:13  
Check for newer version at http://dev.w3.org/cvsweb/~checkout~/2002/scribe/ Guessing input format: RRSAgent_Text_Format (score 1.00) Succeeded: s/ f2f5/ f2f6/ Succeeded: s/Sandor/Sandro/ Succeeded: s/+1// Succeeded: s/"unifies"/does not unify/ Succeeded: s/named classes/named graphs/ Succeeded: s/ot fing/to find/ Succeeded: s/wee/need/ Succeeded: s/need/needed/ Succeeded: s/to RDF/of RDF/ Succeeded: s/reflesive/reflexive/ Succeeded: s/beacuse/because/ Succeeded: s/not a big price to pay/is not sufficient justification for introducing a classification vocabulary in RIF/ Succeeded: s/diff between what is proposed and RDFS subclassing/question/ Found Scribe: Hassan Found ScribeNick: Hassan Default Present: +39.047.101.aaaa, ChrisW, Sandro, Hassan_Ait-Kaci, Dave_Reynolds, StellaMitchell, LeoraMorgenstern, josb, Harold, Gary_Hallmark, Allen_Ginsberg, AxelPolleres, IgorMozetic, luis_polo, MichaelKifer Present: +39.047.101.aaaa ChrisW Sandro Hassan_Ait-Kaci Dave_Reynolds StellaMitchell LeoraMorgenstern josb Harold Gary_Hallmark Allen_Ginsberg AxelPolleres IgorMozetic luis_polo MichaelKifer Regrets: Fran´┐ŻoisBry PaulaLaviniaPatranjan JeffPan MichaelSintek Agenda: http://lists.w3.org/Archives/Public/public-rif-wg/2007Jul/0188.html Got date from IRC log name: 31 Jul 2007 Guessing minutes URL: http://www.w3.org/2007/07/31-rif-minutes.html People with action items: WARNING: Input appears to use implicit continuation lines. You may need the "-implicitContinuations" option.
[End of scribe.perl diagnostic output]