W3C

- DRAFT -

RIF Telecon September 11 2007

September 11 2007

Agenda

See also: IRC log

Attendees

Present
Adrian Giurca (agiurca), Christian de Sainte-Marie (csma), Dave Reynolds, Deborah Nichols, Doug Lenat, Gary Hallmark, Harold Boley, Hassan Ait-Kaci, Igor Mozetic, Jos de Bruijn, Leora Morgenstern, Michael Kifer, Paul Vincent, Paula-Lavinia Patranjan, Sandro Hawke, Stella Mitchell
Regrets
Allen Ginsberg, Axel Polleres (travelling, not sure whether I can dial in)
Chair
Christian de Sainte-Marie
Scribe
Paul Vincent, Stella Mitchell

Contents


Admin

 

<agiurca> Next week I will scribe

Christian: Action-295 on Deborah N is continued

DeborahN: Also, I have the June 26th minutes ready to send. I just have to ask ChrisW about something first.

Christian: Action-324 obsolete

Christian: Any objections to accepting last week's minutes?... none, Sept 4th minutes are accepted.

Liason

PRR liaison: no news except submission at OMG before next F2F

Sandro: Re: Liaison with OWL WG (for OWL 1.1). Membership in that WG should be settled down by November. But if someone in this group wants to join and be the liason we can settle on that now

JosB: XML schema WG was queried by me, but no response

Christian: Action-399 is closed

<sandro> ACTION: Sandro to find out from XML Schema WG's staff contact how we should proceed with getting a response to Jos' email [recorded in http://www.w3.org/2007/09/11-rif-minutes.html#action01]

<rifbot> Created ACTION-342 - Find out from XML Schema WG\'s staff contact how we should proceed with getting a response to Jos\' email [on Sandro Hawke - due 2007-09-18].

F2F

<sandro> http://www.w3.org/2005/rules/wg/wiki/F2F7

Christian: No open actions on F2F7. All RIF WG members should register to indicate whether or not they will attend

<sandro> F2F7 Registration/Regrets Form

Christian: F2F7 objectives: to publish BLD as early as possible in October, so issues must be settled as much as possible.
...and BLD XML schema to be decided
...freeze BLD version ASAP and email link - action for Harold

<sandro> ACTION: Harold to freeze and editors draft of BLD when he's ready (soon), and send the WG e-mail with the frozen version (or a pointer to it). [recorded in http://www.w3.org/2007/09/11-rif-minutes.html#action02]

<rifbot> Created ACTION-343 - Freeze and editors draft of BLD when he\'s ready (soon), and send the WG e-mail with the frozen version (or a pointer to it). [on Harold Boley - due 2007-09-18].

Christian: After the BLD is frozen, please review it and send any issues you have with the draft to the chairs

Sandro:Let's try to minimize surprise issues at the F2F; think about and raise any important issues you have beforehand

Christian: We are aiming to have a frozen BLD draft by this Friday, September 14th

Christian: Possibility of F2F8 at tech plenary in early November in Boston, what's the feeling of the group?

Sandro: The reason that I think a F2F in November is a good idea is because it will be a crucial time - the time when we will be making a case for extending the life of the working group.

Christian: Poll in F2F8 on November 5-6 in Boston?

<PaulVincent> +1 also to F2F8 in November

<sandro> +1 to F2F8 Nov 5-6 in Boston

<DaveReynolds> -1 (I would not be likely to make it)

<Harold> I try to come.

<josb> not yet sure

<csma> +1 to F2F8 in Boston

<PaulaP> I also try to be there

<agiurca> I also try to be there

<IgorMozetic> +1 to F2F8 in Boston

<GaryHallmark> +1 to f2f8

Christian: W3C rule that you must decide on F2F at least 8 weeks before holding one. That's why it is important that we decide ASAP

Sandro: We need ChrisW's input to make the decision

Christian: We will discuss at the chair's meeting and let the WG know

UCR - UC8 worked example

<GaryHallmark> we already discussed bpel orchestration (UC9), last week

DaveReynolds:We can discuss mine, UC8

Christian: Any objection to discuss UC8, although it wasn't on agenda?... none

<Harold> http://www.w3.org/2005/rules/wg/wiki/UC8_Worked_Example

DaveReynolds:The link is above. I wrote this a long time ago, so some syntax and issues are out of date
...this UC is about vocabulary mapping - typical use of rules with RDF
...the rules are simple: take triple patterns, and deduce a new type or new set of values in the target ontology
...I wrote the rules in Jena Rules and then did the analysis
...Re: issues. the rules are mostly Horn (some syntactic sugar about conclusions in head),so not so many issues
...quantification over RDF predicates - with frames there is now no restriction on quantifying over RDF predicates, so this is no longer an issue
...datatypes: also resolved
...builtins: we still need some more, but shouldn't be too controversial
...bNodes: in examples like this (which are realistic), people are treating bNodes as skolem constants
...would need gensym equivalent to do what Jena rules does
...metadata: I had based my example on my proposal at that time,but that's not the way we're going now
...xml syntax doesn't match what we have right now, but it won't be hard to redo once the XML syntax is solidified

<Harold> Re builtins: http://www.w3.org/2005/rules/wg/wiki/List_of_functions_and_operators

Sandro: Did you do this by hand, or automate?

DaveReynolds: Largely hand editied, but some generated

Sandro:Do ou have a sense of how hard it would be to automate the translation?

DaveReynolds: Handling the covered items wouldn't be too difficult, but there is much that isn't covered by RIF

Christian: Rule where condition would check against RDF data set, and modify the data set?

Sandro: It would be nice to go between N3 and Jena Rules by November as a demonstration

DaveReynolds: That would be difficult for me to do by then

Sandro: Is there anyone else you work with who could do it?

DaveReynolds: Maybe

Naming conventions

Sandro: I started with a strawman.

<sandro> Proposed Naming Conventions

Sandro: Naming conventions make it easier for everyone to work with a vocabulary
... easier for users, and easier for the people coming up with new names
... In this proposal, I followed the conventions of Java
... CamelCase for class names, and don't abbreviate, and use noun phrases
... property names are more controversial: start with lowercase, and other than first letter, use CamelCase, and use noun phrases
... and, convention in industry is to keep them singular
... also, I propose that we don't use all caps for any names

Christian: Any questions on this proposal?

DaveReynolds: I think this is useful, but not critical, and I'm happy with this proposal, but I would add one thing ( something about clarifying names that can be ambiguous)

<josb> +1 to proposal Sandro and suggestion daver

<agiurca> we notice this to the WG for a long time. +1 to Sandro proposal

<Harold> We had a WG decision to rename Con into Const.

<Harold> In our fully striped XML syntax we followed the Java convention.

Christian: So, you are agreeing with Sandro?

<Harold> no.

<Harold> only the basic java convention.

Christian: I would like to comment that complete phrases can result in excessively long names for classes and properties

<GaryHallmark> then compress it

Harold: Things are never completely unambiguous anyways

Sandro: I see your point

Christian: I think UniversalTerm is much more clear than Uniterm

<Harold> Christian, yes we dont want to expand Uniterm into UniversalTerm

<Harold> since UniversalTerm is *still* ambigous

<sandro> Maybe amend: If an abbreviated term is no more confusing or misleading than a longer term -- without external explanation -- to the target audience, then it may be used.

Christian: Let's have a resolution about this

<Hassan> I second Sandro's point

Harold: In BLD we are only following Java conventions; not everything Sandro proposed

Sandro: No, you don't completely follow Java conventions - e.g. you don't use noun phrases
... I strenuously object to not using noun phrases in our naming conventions

Sandro: Settling on good naming conventions is critical for a usable exchange format

<GaryHallmark> +1 naming conventions

<sandro> PROPOSED: we will have naming conventions; people edit the page to propose theirs.

Christian: Any objections to above proposal? If you modify or object to someone else's proposal say what is your reason.

<sandro> PROPOSED: we will have naming conventions; people edit the page to propose theirs (with explanation and reasons for any differences from what's already on page)

Christian: Any objection to above? ... none

<sandro> RESOLVED: we will have naming conventions; people edit the page to propose theirs (with explanation and reasons for any differences from what's already on page)

RDF in BLD

ill-typed literals

<DaveReynolds> I thought I voiced some reservation

Christian: Re: ill-typed literals, do you have a counter proposal, Dave?

DaveReynolds: It has to do whether this applies to embedding or combined model case
... it's OK in the embedding case, but not in combined model case
... I think there should be a flag that says whether it's OK to let an ill-typed literal through the translation

Christian: In the combined model situation, it doesn't make sense because you would not translate RDF graph into rules.

DaveReynolds: Exactly

JosB: In the combined model case, you might encounter this type of thing if you query
... but you would not have it in the rules themselves

Christian: You would query to check entailment of the condition, and would never have that with ill-typed literal

JosB: No, one could write any type of query. There might be variables in it and one of the variables substitutions could be an ill-typed literal

DaveReynolds: e.g. a rule that queries an RDF graph to query type of literal and the literal is ill-typed)
... I want to be able to query an RDF graph in its native form, and if my system can handle ill-typed literals, then I want to see them

JosB: We need syntactic correspondence between symbols in RDF and symbols in RIF

Christian: If RIF is used only for interchange, then this is not an issue; there is an issue when embed an RDF graph in a rule set, but not when you have rules that are about RDF graphs

JosB: There are still entailments in that case

Sandro: As an example, you could have a rule that says ' if x works for ILog then x works in France'.
... if x is an ill-typed literal, then the conclusion will have one

Christian: In RIF, you will never have an instance of that because it will be translated to a rule language before being applied to the RDF data
... RIF has an entailment relation for the purpose of telling you how to translate, so that you can preseve the entailment relation

Sandro: I see, you are saying that that aspect should be left up to the implementation

JosB: I don't understand - how can you specify part of entailment relation and not another part? Entailment either holds or doesn't hold

DaveReynolds: (something about well-formed document)

Christian: My point is that it is specified in RDF, and the RIF semantics doesn't have to handle that case

Hassan: I strongly support Christian, and also what MichaelK has been advocating
... making the combined models normative is not a good idea

Christian: I'm confused: I though in the combined model, we do not care about ill-typed literals
... but in the embedding case (which MichaelK supports) we would have to deal with ill-typed literal?

Hassan: I'm not sure about ill-typed. But, the semantics of RDF is not relevant for our RIF purposes

Christian: I would like to ask if others are confused about where ill-typed is an issue?

DaveReynolds: Hassan is addressing the question of embedding vs. combined model, not specifically ill-typed

Christian: Am I right that ill-typed is an issue in the embedding and not in combined models?

JosB: It is an issue in both cases

Christian: I agree with Dave then, that a flag (for whether ill-typed literals are OK) is a good idea

Christian: Any other comments on this topic?

<Harold> As I mentioned in the previous telecon, handling ill-typed literals need to be dealt with at least in the (partial) interoperability part of RIF.

entailment regimes (embedding vs. combined models)

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

Christian: Dave, regarding your concern about accessing RDF data: do you see in the normative section (Semantics, Common Interpretations) where it talks about correspondence between RDF triples and RIF molecules?

Dave Reynolds: OK, I hadn't seen that

DaveReynolds: But still, I think bNode discussion is only in informative part and I'm not comfortable with that

JosB: Re: bNodes the embedding is only used for reasoning, and when you reason you can skolemize existentially quantified varialbes
... if you want to use them for representation and not reasoning, then you are deviating from the semantics of bNodes

DaveReynolds: But we need rules that operate over RDF data and those rules will need builtins, such as SPARQL has, to test various things about the data

AOB

Christian: Frozen BLD draft will be available by Friday, September 14th.

Christian: Any support for adjourning?

<PaulaP> +1

<Hassan> +1

<PaulaP> bye

<PaulVincent> bye

<DougL> bye

<agiurca> bye

Summary of Action Items

[NEW] ACTION: Harold to freeze and editors draft of BLD when he's ready (soon), and send the WG e-mail with the frozen version (or a pointer to it). [recorded in http://www.w3.org/2007/09/11-rif-minutes.html#action02]
[NEW] ACTION: Sandro to find out from XML Schema WG's staff contact how we should proceed with getting a response to Jos' email [recorded in http://www.w3.org/2007/09/11-rif-minutes.html#action01]
 
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.128 (CVS log)
$Date: 2007/09/11 16:32:37 $