RIF Telecon 28 August 2007 -- 28 Aug 2007

<ChrisW> Scribe: Paula Patranjan

<ChrisW> scribenick: PaulaP

Admin

<ChrisW> http://lists.w3.org/Archives/Public/public-rif-wg/2007Aug/att-0047/21-Aug-07-rif-minutes.html

ChrisW: we have minutes from last week's telecon
... any objections?
... minutes accepted
... action review
... Ben Grosof not able to perform his actions

so, better drop his actions

Liaison

ChrisW: move to Liaison - nothing to report

F2F

ChrisW: anything to report?
... there is a registration form for the next F2F

<ChrisW> F2F7 reg: http://www.w3.org/2002/09/wbs/38457/f2f7reg/

it is also linked from the wiki

Sandro: please fill out the form even if you are not coming

ChrisW: we'll have coffee breaks, lunches, and a dinner
... dinner either at a local restaurant or in NY
... we could do a social event on Friday

<AxelPolleres> :-)

ChrisW: questions on F2F?
... comments?

UCR

ChrisW: last week we had a discussion on one of the use case examples

Hassan: phone connection at F2F?

ChrisW: as usual for the telecons
... we discussed the use case example of Leora

<Harold> http://www.w3.org/2005/rules/wg/wiki/D._Evaluation%3A_Use_Cases

ChrisW: use case 6
... does anyone want to discuss one of the examples?

ChrisW lists the existing use case examples

ChrisW: how about use case 10?
... is the author of this use case example here?

<sandro> AxelPolleres, you here?

Axel: better to discuss it next week

ChrisW: use case 9 by Gary

Gary: this one uses production rules

and a simple XML Schema

<AxelPolleres> I prefer to go over it again wrt. the latest changes in BLD, etc.

Gary: production rules increment a score counter
... I changed them to an assert
... we need aggregation here
... extended with an aggregate operator
... the issue is that still requires operators like sum for working with the working memory

<Harold> Gary, this looks similar to Mark's MISMO efforts: http://lists.w3.org/Archives/Public/public-rif-wg/2007Jun/0016.html

Gary: the order of conjuncts becomes important
... it doesn't really translate production rules
... we have here deduction rules with aggregation
... sum into a score
... like a Prolog findall

Sandro: one possibility is just to add the scores
... another possibility is to explicitly aggregate them
... by using a kind of 'big rule'

<ChrisW> acl sandro

<Zakim> sandro, you wanted to ask about non-aggregator way to do this in BLD

Harold: I agree with Sandro's point

Gary: need for actions to update global counters
... but not the case for this particular example

<sandro> Sandro: that is, it's possible to do this without aggregators, if you're willing to maintain pointers to all the rule results. Then it's BLD, but it's harder to maintain.

Allen: you say that there is a set of increments
... does RIF really handle this kind of example?

Gary: using forall instead of existential would not change the effect of rule

Doug: reflection is used in rule languages for such kind of examples

ChrisW: we don't have this capability in the first version of RIF

<AxelPolleres> couldn't what doug says be "emulated" by a kinda trigger-flag per rule?

ChrisW: other comments on Gary's example?

<dougL> good idea, axel; i think that would work.

Sandro: no aggregation and assert in the current BLD proposal

BLD - RDF in BLD

ChrisW: continue discussion of use case examples next week

<josb> yes

ChrisW: review of actions

<ChrisW> http://www.w3.org/2005/rules/wg/wiki/Core/RIF-RDF_Compatibility

Jos: not sure about the right place for the RIF RDF compatibility

Harold: it is the right place
... and it shows that we need same kind of work for OWL compatibility

<AxelPolleres> Do we liaise with OWL 1.1? sorry might have missed that, if mentioned earlier in the call before I entered.

Jos: I can also outline some issues for the OWL compatibility

BLD - XML Syntax

ChrisW: Sandro has released some work on XML syntax

Sandro: I proposed a strawman XML syntax a few weeks ago

<sandro> Example RIF Instance

Sandro: a complete example was also given

<DaveReynolds> Yes, I can see it

<Hassan> +1

<AllenGinsberg> got it

Sandro: I discovered a couple of issues

ChrisW: take us through a part of the example

<ChrisW> action chris to put action 330 on the agenda next week

Sandro: one of the issue is the usage of contants

<ChrisW> rifbot, only my mother calls me christopher

Sandro: different possibilities exist to specify e.g. global and local constants

<sandro> (modified) BLD Abstract Syntax

Sandro: can be specified
... global constant has to be a URI
... questions about that?

Harold: we have an XML syntax already
... use of attributes for the difference between global and local constants
... so no need for different types of tags for the different kind of constants
... also other controversial features in the proposed syntax
... in XML we have the order of the child elements

Sandro: but then there might be a problem when the order is unimportant
... how does a parser know if it is ordered or not?

Harold: inside uniterms the order is important
... by default in XML you keep the order of child elements
... e.g. the order of slots does not matter

Michael: a few weeks ago I asked Sandro to explain what is wrong with the current, proposed XML syntax and why is the new proposal better

Dave: is there a formal definition of the first proposed XML syntax?

Sandro: it was just a placeholder
... and I thought the first XML syntax was not a complete proposal

Hassan: can it be reconciled, Sandro? can you adapt the asn to match the official version?

Sandro: there are things that I don't agree with
... the question is whether to refine them or discuss them first

Hassan: if there is no agreement, then we should discuss the open issues regarding syntax

ChrisW: the XML syntax in the BLD draft has no official status
... not clear where to go from here

<DaveReynolds> parsetype collection is not controversial in RDF

Harold: RDF as a top level, controversial features in RDF are examples of issues that are problematic in Sandro's proposal
... it also deviates from our use cases
... we can use the abstract syntax to generate directly the XML syntax
... so no need to invent a new language

Sandro: I didn't generate a new language

Harold: but you changed the abstract syntax

Sandro: yes, this I did

<sandro> class Ruleset

<sandro> property formula : Rule*

Sandro: this is the change I did

Harold: attributes tell you about the order

ChrisW: we don't have an XML syntax
... the existing one has no official status
... Harold, is your main objection the fact that the new proposal is a RDF syntax?

Harold: even in the RDF community some of the issues are controversial
... one main point is that in RDF you can have these open properties and this is wrong for a syntax
... refinement of constants not in the semantics

<DaveReynolds> Global v. local constants are in the semantics already surely?

Harold: I will send an email with all my objections to the new proposal

<Harold> DaveR, we have a formal abstract syntax and an EBNF definition.

<AxelPolleres> I have at least a worry here: the RDF syntax already had some complicating effects on OWL, IMO, my personal opinion is at least that I am unsure whether its a good idea to have OWL in RDF... I would likewise be worried for RDF in RIF.

Sandro: I would like to hear concrete objections

<Harold> "The primary normative syntax of the language must be an XML syntax." http://www.w3.org/2005/rules/wg/charter.html#xml-syntax

Michael: the first proposal is a proposal and shouldn't be ignored

ChrisW: it is a valid alternative

<sandro> Harold, this IS AN XML SYNTAX. It has an XML Schema.

Dave: it would be good to have a pointer to what this alternative is
... among the concrete issues was how we could carry metadata
... so we have to include a solution to this issue

<Harold> DaveR just argues for an RDF SYNTAX, not AN XML SYNTAX, because of metadata for rules.

<AxelPolleres> An example why I worry is that OWL/RDF splits one OWL statement to a set of triples, however a subset of those triples usually makes no sense, whereas in RDF alone a subset of triples makes perfect sense. So, on the RDF level, you cannot decide whether an OWL/RDF graph reflects a valid set of OWL statements.

<DaveReynolds> Harold - Sandro's proposal is an XML syntax which happens to be largely RDF compatible

<AxelPolleres> We would run into the same problem in RIF.

<josb> pieces of RDF could be embedded an the RIF XML syntax

Sandro: we have to give people the chance to look into issues before taking decisions

<sandro> Issue: ordering/structure encoded in instance

<sandro> Issue: how to serialize and model datatype literal values

<sandro> Issue: how to serialize and model local and global constants

Sandro: RDF vs non-RDF syntax is an issue

<josb> Axel, "So, on the RDF level, you cannot decide whether an OWL/RDF graph reflects a valid set of OWL statements." This is not true. One can guess an OWL ontology, and check whether it represents this ontology

Sandro: we could figure out what we need by means of use cases

Axel: concerns on using RDF for the syntax
... you mix the semantics of the data and the semantics of the structure
... I would prefer a kind of separation

Sandro: the XML Schema is not for RDF but for RIF

Axel: I see
... still...

ChrisW: I don't see the worry here

<josb> in that case, using the RDF namespace is a bad idea

Axel: would this data would be used as RDF somewhere?

Sandro: take for example Prolog
... for RIF would be in fallback processing

<Harold> In the current syntax, we have unordered rulesets as the default, so <Ruleset> . . . </Ruleset> means <Ruleset ordered="no"> . . . </Ruleset>.

<josb> I have already heard from people who want to query RDF stores for rules about certain resources

<AxelPolleres> sandro: might make perfect sense, good point, I do not object upfront, just wanted to make it clearer for me.

ChrisW: there is a need to clarify the difference between the two existing proposals for syntax

<AxelPolleres> yes, agree, that there are several issues around this, we should collect them in one place.

<Harold> So, the unordered/ordered distinction is very natural to express via an XML attribute, and defaults handled via XML's attribute defaulting mechanism.

<Harold> We dont need RDF's rdf:parseType="Collection".

ChrisW: better to break this down issue by issue but we need to start somewhere

<sandro> is there an important difference between whether you say ordered="yes" or rdf:parseType="Collection"

BLD - Planning for next WD

ChrisW: concrete plans for the next BLD working draft
... there were several items I wanted to have in the next draft
... one of them was the XML syntax
... seems we have consensus on the RDF compatibility
... the issue of classification is another one
... I made a proposal a couple of days ago
... what do you think about this plan?

Michael: I propose that just the second part of the RDF compatibility section should go into the next BLD draft
... I don't see the need for a combined language, just for the embedding
... I sent an email on that

ChrisW: so maybe we are not that close to consensus as I thought
... we'll try to discuss it next week
... what is the timeline we would need to get the next draft out?

Michael: it depends on the form we'll have for the draft
... I propose to split the document in two
... one document on the framework
... and the actual dialect will be explained in more simpler terms
... but there is some work to spit the document in two

<sandro> +1 splitting document, so BLD is specified in a smaller/simpler document

Harold: I suggest we keep the part on the architecture instead
... and have a draft before the next F2F

Michael: I didn't think at the Arch part when talking about the framework

ChrisW: how long will it take?

Michael: 2 to 3 weeks to split the document

Sandro: will the framework also be a standard then?

Michael: yes, but not part of a particular dialect

<josb> in that case, it has to be a standard

Michael: I would do it before the F2F

ChrisW: any other issues regarding the work that needs to be done?

Dave: slotted terms and formulae should be discussed
... it was not clear why we need this

<MichaelKifer> here is the msg where I raised doubts about one part of the RDF+RIF document: http://lists.w3.org/Archives/Public/public-rif-wg/2007Aug/0039.html

Harold: controversial issues should be postponed for the 3rd draft

BLD - Frames/Classification

ChrisW: I proposed that we include the frames and classification section in the next draft
... with label under discussion together with pros and cons

Dave: I have a couple of issues with it
... first the group should take a decision on it
... data model issue
... these issues should first be discussed in the group before publishing

Michael: useful not only for data model interchange
... also for optimizations

<sandro> +1 keeping the text, labeling as under discussion

ChrisW: straw poll on keeping the text and labelling with under discussion

<DaveReynolds> - 0.5

<Harold> +1

<dougL> +1

<StellaMitchell> +1

<AxelPolleres> +1

<josb> 0

<Hassan> 0

<AllenGinsberg> 0

<PaulVincent> 0

<MichaelKifer> +1

Dave: I guess I could live with it in the draft, but not in the last call

<GaryHallmark> +1

<luis_polo> 0

ChrisW: resolution

<ChrisW> PROPOSED: To include the section on frames/classification in the next BLD WD, suitably labelled as "under discussion" with rationale for/against

<ChrisW> RESOLVED: To include the section on frames/classification in the next BLD WD, suitably labelled as "under discussion" with rationale for/against

ChrisW: this doesn't mean this is closed
... AOB?

- DRAFT -

RIF Telecon 28 August 2007

28 Aug 2007

Attendees

Contents