See also: IRC log
<ChrisW> sandro, u there?
<sandro> I'm here, Chris.
<sandro> (stuck at home because my car battery died, but here. :-)
Chris: Next meeting is next Tuesday, as usual.
<scribe> ACTION: review to [recorded in http://www.w3.org/2007/04/17-rif-minutes.html#action01]
<sandro> scribenick: LeoraMorgenstern
Powder working group: representative not present
Ontology group: has started mapping between sBVR and OWL/RDF
<AlexKozlenkov> betweeen SBVR and OWL/RDF
Chris: Regarding the Core document ...
Chris: Not much public feedback
yet on Core document, beyond Jeremy's email regarding typos
... Everyone in working group should be announcing release of first working draft to colleagues, should be sent to mailing list, etc.
... If anyone knows of appropriate place to announce release of working draft, we should send it there.
Sandro: There's a Wiki page called Outreach, where people should list organizations to which to send documents.
Chris: This is a good record to remember what was done last time wrt announcements. Also helps prevent overlap [of announcements sent to same groups by different people].
<Harold> Perhaps we should combine the WD1 announcement with a reminder: Please Comment By 27 April 2007
Sandro: We have less than 10 days left to the comment period.
Chris: We will still accept comments beyond that date, but from a process point of view, we are not obliged to respod to such comments.
Chris: There was some discussion of this last week, and there's been email follow-up.
Chris: Michael, Harold, and Dave, can you summarize the discussion? Are we near consensus yet?
Harold: Summary: everyone is fine
... my own opinion: in the unlikely event that the web changes from URIs, all groups will need to change.
<Harold> I think we can call it rif:iri
Chris: Is there anyone who disagrees that we should use IRIs instead of URIs in the spec?
Francois: IRIs shouldn't be referred to as resources, but as pointers.
<sandro> PROPOSED: We'll use IRIs
Chris: proposing to close Issue
30 by agreeing to use IRIs instead of URIs everywhere in core
... Is anyone opposed?
... Hearing none, we've closed issue 30.
... Naming: should we change naming from rif:uri to rif:iri?
Harold: yes to Chris's suggestion
Michael: Maybe should be rif:id
Harold: But in XML, id has another meaning. Any [suggested] term might have another meaning.
Sandro: Are we naming things by serialization or by thing identified by serialization?
Chris: Michael, are you objecting, or just sounding out alternatives?
Michael: no objection, just trying to find a neutral term.
Chris: In interest of expedience, I suggest accepting rif:iri; we can always change if something better comes along.
Hassan: Want to be able to adapt to whatever the standard is. That's what Michael is aiming for.
Chris: so what would the actual name be?
<Harold> Hassan, but such a level of indirection could later diverge to another kind of URL/URI/IRI/..., which don't want to do in RIF.
<DaveReynolds> fine by me
<MichaelKifer> rif:gid (global id) or rif:rid (resource id)
<sandro> PROPOSED: Change it to rif:iri for now, we can discuss level of indirection and change it later if there's consensus on a new name
<Harold> +1 to Chris
<sandro> RESOLVED: Change it to rif:iri for now, we can discuss level of indirection and change it later if there's consensus on a new name
Proposal is resolved.
<sandro> ACTION: mkifer to change URI to IRI and rif:uri to rif:iri [recorded in http://www.w3.org/2007/04/17-rif-minutes.html#action02]
<rifbot> Created ACTION-275 - Change URI to IRI and rif:uri to rif:iri [on Michael Kifer - due 2007-04-24].
ACTION on Michael Kifer to update spec to reflect both resolutions:
Dave: absolute, relative addressing, stuff from SPARQL, Dave will paste into IRC
<DaveReynolds> The paragraph was in: http://lists.w3.org/Archives/Public/public-rif-wg/2007Mar/0133.html
<DaveReynolds> The para was ...
<DaveReynolds> In the concrete XML and human readable syntax relative IRI references
<DaveReynolds> are permitted in which case they will be resolved relative to a base IRI
<DaveReynolds> as per Uniform Resource Identifier (URI): Generic Syntax [RFC3986] using
<DaveReynolds> only the basic algorithm in Section 5.2. Neither Syntax-Based
<DaveReynolds> Normalization nor Scheme-Based Normalization (described in sections
<DaveReynolds> 6.2.2 and 6.2.3 of RFC3986) are performed.
Chris: Onto Issue 31
Issue 31: Disjoint Domains for Individuals, Functions, and Predicates
Chris: Anyone thinks we should have disjointness in the Core?
Michael: Jos's objection was that
this would make compatibility with OWL/DL harder.
... that is, not having the assumption of disjointness would make achieving compatibility harder
Chris: OWL is talking about
changing this in OWL 1.1, but realistically, this is more than
a year away.
... Concern about making interoperability more difficult --- makes translation more difficult.
Hassan: I would think just the opposite --- the more restrictions, the harder
Chris: No good reason technically
or mathematically to have this restriction.
... Nevertheless, there are many systems that make this simplification.
<Francois> +1with Hassan
Hassan: It's not a simplification, it's a complication!
Sandro: What about the logic textbooks that I've seen that make this assumption?
Hassan: It's not necessary, and it's silly!
<sandro> Hassan isn't actually saying it's not commonly done -- he's disagreeing but saying something else.
Michael: Useful, possibly, in a textbook, but useless in a language. There are didactic reasons for the disjointness assumption, but no computational reasons.
Francois: It's a 19th century tradition.
<sandro> Francois: it's a 19th century tradition --- like in Fortran where all variables that start with an n (i) are integerrs, etc.
<sandro> Francois: There's no reason to do it any more.
<Hassan> Good explanation Francois !
Sandro: No one is arguing whether
you can tell the type of symbol by looking at it.
... what they are arguing about is whether the same symbol can be used as individual and predicate in same formula.
<Hassan> Yes - Java , C, C++ do it !
Sandro: No one is going to the extreme of Fortran
<Francois> +1 with Hassan's explanation
Hassan: Francois was not suggesting a direct analogy; he was pointing out the silliness of this particular convention
<sandro> but it's a different convention.
Hassan: and arguing, that just as the Fortran convention can be dropped, so can this convention.
<Francois> Chris' comment is not fair!
Chris: Are we so passionately convinced that this is a silly convention that we will tell people not to adhere to these conventions even though that will make their translations harder?
<AlexKozlenkov> massive noise
Hassan: we are not authorizing or forbidding anything. [i.e., they can, if they wish, adhere to the convention of disjointness.]
Hassan: No problem in Java, e.g, having foo(foo), where first instance is a function, and second is an individual.
Hassan: We should have a better reason for a design choice than, "the neighbors do it."
<sandro> forall x x(x) <---- you want to allow that, Hassan?
Francois: reiterates that we should not still use the style of the 19th century at the beginning of the 21rst century.
<sandro> DaveReynolds: Could the people who are opposd to DS say whether they want ONDS (OWL1.1) or OS (RDF)?
<Hassan> Sandro : I do not want to allow anything meaningless, I do not want to forbid something meaningful
<sandro> +1 DaveReynolds
Dave: Are those who are opposed to DS (disjoint sorts) supporting ONDS (punning: Overlapping names, disjoint sorts) or OS (overlapping sorts)?
Chris: believed that Francois supported ONDS
<Hassan> +1 with Francois - distinct categories should be an *option* not a requirement
<sandro> Francois: it seems to me that what is being called "sorts" here should be called "syntactic categories"
Chris: We may not be able to push
much further with this. However, people aren't really
addressing the question: Does it make sense to make the
translation more difficult?
... Perhaps we need more examples of translations so that this discussion can become more real.
Sandro: Is anyone opposed to OS?
<Francois> What is OS?
<PaulaP> overlapping sorts
OS = overlapping sorts: same symbol can denote any number of things.
<Harold> OS: Overlapping sorts. Symbols uniquely denote semantic objects but these objects themselves can be more than one thing (a function and/or a predicate and/or an individual). This is the approach used in common logic. A common misconception is that this is necessarily second order, but as shown in my logic textbook (Enderton), it is still first order.
Chris: hesitant to make decision. What would Gary, Paul Vicent, Jos say?
<AlexKozlenkov> Hassan +1
Hassan: I don't understand the OS thing... The description makes it sound as if we're talking about semantic objects, but this is just a discussion of syntax.
Hassan: OS description doesn't make sense to me.
Chris: Difference between OS and
ONDS: ONDS --- can use the same symbol syntactically, but depending
on its position, it can be different things (predicate, function,
... OS:unique symbol denotes object which can be used as predicate, function, individual
Chris: admits that perhaps the description could use clarification
Dave: The difference is somewhat akin
to the difference between OWL full and OWL DL.
In OWL full : if you say x and y are equivalent, it the case both that individual classes x and y have same extensions and that predicates x and y have the same extensions.
Dave: but in OWL DL: even if individual classes x and y are equivalent, there's no claim about equivalent extensions for predicates x and y
Chris: (elaborating on Dave's example): so you can say P=Q for individuals, but doesn't say anything about the predicates P and Q being equivalent.
Chris: The OWL 1.1 spec has a good
explanation of this issue.
... people either seem to be ambivalent or expressing support for OS approach.
<AlexKozlenkov> We probably need a good test case
Sandro: there appears to be much confusion about this issue, so we need to look at test cases.
Chris: Would be good to gather
some examples, and especially what this means for
... Any volunteers for examples of ONDS translation into both OS and DS?
<Harold> Should we use this occasion to liaise/work with colleagues from OWL 1.1?
Chris: Hearing no volunteers, will send out email msg to recruit volunteers
ACTION on Chris to recruit volunteers
<sandro> ACTION: ChrisW to recruit someone to write examples of ONDS transation [recorded in http://www.w3.org/2007/04/17-rif-minutes.html#action03]
<rifbot> Sorry, couldn't find user - ChrisW
<sandro> ACTION: Chris to recruit someone to write examples of OS/ONDS translations [recorded in http://www.w3.org/2007/04/17-rif-minutes.html#action04]
<rifbot> Created ACTION-276 - Recruit someone to write examples of OS/ONDS translations [on Chris Menzel - due 2007-04-24].
Chris: Discussion Issue 25, compatibility with RDF in RIF corehttp://www.w3.org/2005/rules/wg/track/issues/25
<sandro> Harold: Object-centered syntax object[property->>value]
<AlexKozlenkov> it is right
Chris: can write rules conditioned on properties.
Dave: basic notion of triple predicate with some syntactic sugar seems very reasonable.
Dave: Jos's paper also showed how entailment rules would be handled.
Michael: I think this can be added to the core.
Chris: Anyone hesitant/skeptical about this approach?
Sandro: Concern raised in email: would like to have binary predicate approach available too, if possible.
<AlexKozlenkov> variables over predicate smay be good to have
Michael: Not when using variables over predicates.
Sandro: but in the 80% of cases where you don't have that ...
<IgorMozetic> Sandro: why would you like binary predicates?
Michael: Sandro, so you're proposing to have one approach to cover 80% of cases, and another approach to cover everything?
Sandro: In RIF core, can't
quantify over predicates.
... In RDF rules, need to be able to allow it.
Sandro. Which is why you have this triple predicate. But, is there a way to have the binary predicate option?
<Harold> What about allowing (Hilog-like) 2nd-order syntactic sugar to query a property in property(object,value), semantically equivalent to holds(property,object,value)?
Sandro: Mulling this over, rather than proposing solution.
Michael: This seems more like an implementation issue.
Sandro: Michael is likely right. Will think about this some more.
<sandro> (That is -- Michael says that using __triple doesn't necessarily mean bad performance. EG you can recognize rulesets which don't use predicate quantification.)
<AlexKozlenkov> Why can't we go ahead and do it?
Harold: Suggests: Holds predicate can be used as syntactic sugar for binary predicates.
Chris: Syntactic sugar for binary predicates or all?
<AlexKozlenkov> Any predicates
Harold: This approach can easily be generalized to n-ary predicates.
Michael: Does this mean proposing
variables over predicates in the core?
... (in response to Chris's remark): Orthogonal to discussion of disjointness issues.
<AlexKozlenkov> Example: We receive a message that contains a triple, we find a rule for a matching binary predicate
<AlexKozlenkov> In reactive systems at least, we would exchange terms that then could unify with atoms
Sandro: do you have aggregates? Can you descend into objects themselves and write rules about their slots, etc.?
Chris:Can you query an object for its slots, e.g.?
Hassan: talking about reflection. Yes, JRules can do this.
Sandro: Point was: do we need quantifying over predicates for more than just RDF compatibility? Is it needed in general for RIF?
<AlexKozlenkov> As said before, there are uses for that elsewhere
Chris: A point of clarification: Jos's proposal (translation between RDF and F-Logic) discussed quantification just of RDF properties, not of any predicates. Can we generalize to quantification of predicates in general?
<Harold> Semantically, querying ?property in object[?property->>value] doesnt seem to be easer than querying property(object,value).
It seems to me that once you introduce reification (via Holds), you may as well take advantage of its power.
Harold: <missed it>
Chris: We don't need to support all of RDF in core. Might not have to support quantification over predicates.
<AlexKozlenkov> Sandro: exactly
Sandro: so, could suggest that we use binary predicates, and if want to quantify over predicates, use same extensions as anyone else.
<Hassan> Again - I second Michael's opinion - why forbid?
Michael: But slots also impose the burden. Not necessarily sure disallowing quantification over predicates solves the problem.
<Francois> sorry, I must leave...
<AlexKozlenkov> RDF is just one lead into the issue
Chris: 3 possibilities regarding quantification over variables. (get them)
Michael: Slotted notation is easier to swallow than full quantification over predicates.
Chris: let's go ahead with adding slotted syntax to core.
Alex: is concerned that we are avoiding the issue of quantifying over predicates, but that it will come up again in the future.
Chris: But most systems can't handle quant. over pred., and one does want to have the core reflect what's in most systems.
<Harold> Vendors who dont offer direct handling of predicate variables would need to do the 'holds' transformation into their systems.
Alex and Chris: more discussion about not committing against quant over pred,
<scribe> ACTION: ChrisW to put rifraf on agenda for next week's telecon [recorded in http://www.w3.org/2007/04/17-rif-minutes.html#action05]
<rifbot> Sorry, couldn't find user - ChrisW
<scribe> ACTION: ChrisW to put rifraf on agenda for next week's telecon [recorded in http://www.w3.org/2007/04/17-rif-minutes.html#action06]
<rifbot> Sorry, couldn't find user - ChrisW
<scribe> ACTION: Chris to put rifraf on agenda for next week's telecon [recorded in http://www.w3.org/2007/04/17-rif-minutes.html#action07]
<rifbot> Created ACTION-277 - Put rifraf on agenda for next week\'s telecon [on Chris Menzel - due 2007-04-24].
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/are less/have less/ Succeeded: s/IRI/URI/ Succeeded: s/things/thinks/ Succeeded: s/cates/categories/ Succeeded: s/volunteers?/volunteers/ Succeeded: s/Semantically, querying ?property in object[?property->>value] doesnt seem to be easer property(object,value)/Semantically, querying ?property in object[?property->>value] doesnt seem to be easer than querying property(object,value)./ Found ScribeNick: LeoraMorgenstern Inferring Scribes: LeoraMorgenstern Default Present: Sandro, Dave_Reynolds, ChrisW, PaulaP, Harold, Deborah_Nichols, Francois, Hassan_Ait-Kaci, Leora_Morgenstern, IgorMozetic, johnhall, StellaMitchell, AlexKozlenkov, +43.512.507.aabb, MichaelKifer, [IVML], Mark_Proctor Present: Sandro Dave_Reynolds ChrisW PaulaP Harold Deborah_Nichols Francois Hassan_Ait-Kaci Leora_Morgenstern IgorMozetic johnhall StellaMitchell AlexKozlenkov +43.512.507.aabb MichaelKifer [IVML] Mark_Proctor Got date from IRC log name: 17 Apr 2007 Guessing minutes URL: http://www.w3.org/2007/04/17-rif-minutes.html People with action items: chris chrisw mkifer review WARNING: Input appears to use implicit continuation lines. You may need the "-implicitContinuations" option.[End of scribe.perl diagnostic output]