RIF telecon 14 October 08 -- 14 Oct 2008

<csma> Scribe: Stuart Taylor

<csma> scribenick: StuartTaylor

<csma> agendum+ Admin

<csma> agendum+ Liaison

<csma> agendum+ Public comments

<csma> agendum+ Action review

<csma> agendum+ Action 577

<csma> agendum+ Test Cases

<csma> agendum+ Frames VS objects

<csma> agendum+ AOB (Pick scribe!)

<csma> next item

<csma> PROPOSED: accept minutes of telecon October 7

<csma> RESOLVED: accept minutes of telecon October 7

<csma> PROPOSED: accept minutes of F2F11

<csma> http://www.w3.org/2005/rules/wg/meeting/2008-09-26

<csma> http://www.w3.org/2005/rules/wg/meeting/2008-09-27

<csma> RESOLVED: accept minutes of F2F11

<csma> next item

Public comments

<csma> http://lists.w3.org/Archives/Public/public-rif-comments/2008Oct/0001.html

<csma> next item

StuartTaylor: I will work on some test cases with Yuting.
He will post on the mailing list for feedback next week.

<csma> agendum+ F2F12

F2F12

csma: The preferred date of Jan 15-16 means that Sandro would miss part of F2F12.
Most people would strongly prefer Jan 15-16 rather than 14-15
Is there any objection that F2F12 is on Jan 14-15 ?

<AxelPolleres> I will have difficulties anyway in January, but I think I could join via phone some time at least.

csma: The problem with Jan 14-15 is that people would have to wait until Sat 17th for flights.

<csma> Decision is: F2F12 will be 14-15 Jan 2009

<csma> next item

<ChrisW> ACTION: Gary to start F2F12 wiki page [recorded in http://www.w3.org/2008/10/14-rif-minutes.html#action01]

<trackbot> Created ACTION-621 - Start F2F12 wiki page [on Gary Hallmark - due 2008-10-21].

<csma> http://lists.w3.org/Archives/Public/public-rif-wg/2008Sep/0178.html

<csma> PROPOSED: Core should keep safe disjunction in rule bodies.

<csma> Implementations can be direct or use a well-known preprocessing step.

<csma> Comment: An example of a solution to issue-70 is Option 2,

<csma> which will be able to cope with disjunction. (See

<csma> http://lists.w3.org/Archives/Public/public-rif-wg/2008Sep/0178.html

<csma> and follow up).

Harold: Disjunction and safety discussed in yesterday's Core telecon.

<Harold> http://www.w3.org/2005/rules/wg/track/issues/7

<csma> http://www.w3.org/2005/rules/wg/track/issues/75

csma: Harold, could you explain the link with issue 70?

<AxelPolleres> ISSUE-70... basically both Option 2 and 3 are fine, but 2 is over-cautious.

Harold: Issue 70 is about safeness.
Safeness can be retained using disjunction.

csma: The proposed resolution would be between Option 2 and 3 from Axel's email?

<AxelPolleres> Option 3 in the original mail has a typo in bullet 3 ( Or instead of And)

<Harold> http://www.w3.org/2005/rules/wg/track/issues/48

<Harold> PROPOSED: RIF Core will include member (#) but syntactically

<Harold> restricted its use in rule conditions.

<Harold> Rationale 1: PRD rules almost always start with a member test

<Harold> in the condition.

<Harold> Rationale 2: PR leverages type system from host programming language

<Harold> and that is externally defined and immutable by rules.

<Harold> Comment 1: Note that in RIF-RDF the equivalent property rdf:type

<Harold> would still be permitted in rule conclusions.

<Harold> Comment 2: If PRD introduces member (#) in the conclusion,

<Harold> this restriction to conditions in Core should be reconsidered.

<DaveReynolds> Agreed - we did not take external-# into account in the discussion

csma: This proposed resolution should be discussed in the PRD task force.

<ChrisW> is it possible for something "ground" to not be a "fact"?

<DaveReynolds> A fact is surely just a rule conclusion without a rule condition, at least in Core.

<Gary> forall ?x (A(?x)) is not ground

<ChrisW> but it is a fact?

<DaveReynolds> Part of the conformance clause

<Harold> <Harold> > > http://www.w3.org/2005/rules/wg/track/issues/70

<Harold> <Harold> PROPOSED: Parameterize the conformance clauses of Core with safeness requirements "strict" and "none" (default: "none").

<Harold> <Harold> (modulo nice word for "none")

Harold: We have excluded weak safety and are only considering none or strong.

<csma> next item

Frames VS objects

csma: There is a problem with PRD and the semantics of frames.

<ChrisW> changkai: main difference between frames and objects is multi-valued slots

<csma> o[p->1]

<csma> o[p->1] o[p->2]

csma: The first example above would be translated into PRD as an object that contains p with the value 1.
If the second example is asserted then the value for p would be replaced by 2.
This is not the semantics of frames.

<csma> o[p->(1 2]]

<ChrisW> just extend the syntax to indicate single-valued slots in PRD

MichaelKifer: We need to decide on single or multi-valued slots in core.

<csma> o[p->(1 2)] o[p->3]

<csma> o[p->(1 2 3)] or o[p->((1 2) 3)]?

<DaveReynolds> o[p->{(1 2) 3}] where {} is set not list, no ordering

<csma> I used () for sets, sorry; replace with {} everywhere

<DaveReynolds> csma - ah sorry

<csma> the problem is, if PRD understands only single-valued slots and interprets multi-valued slots as single, collection-typed, values, then, how to differentiate between single values that are collections and collections that represent multiple values?

<ChrisW> just extend the syntax to indicate single-valued slots in PRD

<csma> Frames: o[p->v]

<csma> Object: o[p=v]

<csma> objects are frames of a special kind?

<MichaelKifer> c[p=>person]

<ChrisW> you need a syntax to indicate the difference

<MichaelKifer> c[p {0:1}=>person]

<ChrisW> otherwise you are creating a new schema language

csma: The problem occurs when you import a document from BLD.

MichaelKifer: You would assume that slots are multi-valued.

<ChrisW> we have no "way" to handle cardinality constraints

<MichaelKifer> john[children->bob] and john[children->mary]

<cke> In OO world, we write: john[children->{bob,mary}]

<DaveReynolds> Agree with Michael - Core should support BLD frame semantics (multivalued) otherwise we break the semweb compatibility

<ChrisW> there is no way - "replacement" is non-monotonic

<Gary> java like syntax: class Person [ children=> collection|individual ]

csma: Using a PRD-specific construct will limit interoperability with BLD.

<csma> http://lists.w3.org/Archives/Public/public-rif-wg/2008Oct/0053.html

<DaveReynolds> Agreed - frame slots represent sets of values not bags of values, at least I hope so

<josb> they do (represent sets of values)

<josb> who not just extend the PRD syntax?

<josb> BLD does not need to be extended if something like this is needed in PRD

<Gary> Isn't the notion of frame explained in BLD?

<Gary> ?x=?y :- ?o[single->?x single->?y]

<Harold> What about named-argument terms?

<Harold> A term with named arguments is of the form t(s1->v1 ... sn->vn), where n=0, t ? Const and v1, ..., vn are base terms and s1, ..., sn are pairwise distinct symbols from the set ArgNames.

<josb> o[p=a], o[p=b]

<josb> a=b

<AxelPolleres> aren't slotted *predicates* anyway single-valued... but you don't want that probably either.

<Harold> Axel, I just said that using 'named-argument' instead of 'slotted'.

<AxelPolleres> yeah,but fixed arity maked them probably unusable here.

<josb> yes

<Gary> slotted predicates don't work because we want to change a single slot

<ChrisW> +1 PRD specific solution

<csma> scribe next: Hassan?

- DRAFT -

RIF telecon 14 October 08

14 Oct 2008

Attendees

Contents

Public comments

F2F12

Frames VS objects

Summary of Action Items

Scribe.perl diagnostic output