W3C

RIF Telecon 17 July 07

17 Jul 2007

Agenda

See also: IRC log

Attendees

Present
Harold Boley, Mike Dean, Allen Ginsberg, John Hall, Sandro Hawke, Gary Hallmark, David Hirtle, Michael Kifer, Leora Morgenstern, Igor Mozetic, Jeff Pan, Paula Patranjan, Dave Reynolds, Christian de Saint Marie, Chris Welty
Regrets
Hassan Ait-Kaci, Jos De Bruijn, Axel Polleres, Paul Vincent
Chair
Christian de Sainte Marie
Scribe
Leora Morgenstern

Contents


 

 

1. Admin

Next meeting will be next week, July 24th.

Action 327 on csma to put datasets on agenda for meeting DONE

Action 325 on csm to publish minutes of F2F6 DONE

Action 324 on Chris to ask Deborah Nichols about minutes of 6-26 is still open

Unclear if that was done or not.

PROPOSED: accept meetings of July 10 minutes telecon.

RESOLUTION: accept meetings of July 10 minutes telecon

Action 324 CONTINUED since Chris is not here.

PROPOSED: accept minutes of meetings of F2F6 in Innsbruck

That will be kept open until next week since Harold and possibly others didn't have a chance to go through them.

No amendments to agenda.

2. Liaison

<johnhall> SBVR - nothing to report

No actions for this item.

No news about liaison.

3. F2F7

There are 3 proposals for F2F. One is Harold's, in New Brunswick.

Another is ChrisW's in New York.

There's also a possibility to collocate with OMG meeting in Jacksonville the week of September 26th.

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

John Hall is looking into the possibility of collocating with OMG.

No information yet on costs per person from wiki

csma: if we make the decision next week, earliest for f2f is Sept. 18 [due to mandatory 8-week period between time of decision and f2f].

csma: If we make decision next week, earliest for f2f is Sept. 25.

Chris: We can have f2f In Hawthorne, NY on Sept 27-28 or Oct. 3-4 (W, Th)
... there would be no costs other than transportation

Johnhall: costs [for Jacksonville, FL] would be approx. $120 per head per day

Discussion on dates for possibilities for f2f7 at IBM

ACTION on Leora to find out what dates are available besides the dates that Chris has reserved.

sandro: we should close poll by July 23 and make decision by July 24.

csma: one week for decision, especially in vacation time, is too short.

<PaulaP> no objection

sandro: we've been discussing this for months

chris: agrees with sandro

csma: any objection to closing poll on July 23?
... hearing no objection, we'll close poll on July 23.

4. Technical Design

Action 323 (make a pass at updating the extensibility page based on the discussions and strawmen) on Sandro is due next week

Sandro: would like to put that off for one more week

Action 309-311 (unified strawman proposal for asn --> xml system) are completed; will be discussed today

Action 303, 301 on Jos and Harold, and one on Michael to remove overlapping sorts

Harold: done

Michael: done

csma: any objection to keeping it open until we discuss it?

action 298 on Gary to show how to use xml schema for app data model: completed and closed

Action 260 to put short entry on data sets, outlining the issues on Dave: done

There are a number of actions open in Architecture ( http://www.w3.org/2005/rules/wg/track/products/11 ); these are open until August

<Harold> Modularization: http://www.w3.org/2005/rules/wg/wiki/Arch/RIF_Components/RIF_Dialect_Structure

4.1 Core, Horn, and Production Rules

3 resolutions:

to better understand RIF Core, create 2 task forces,

one focusing on Horn, one focusing on production rules

<csma> PROPOSED: To better understand what RIF Core could be, create two

<csma> task forces in RIF, one focusing on Horn and one focusing on production

<csma> rules.

<csma> PROPOSED: Rename the current "RIF Core" draft: "RIF Horn dialect".

<csma> PROPOSED: Create a "RIF PR dialect".

csma: idea is not to have divergent dialects
... idea is to have a better idea of what a PR dialect would be, what kind of extensions we'd need
... it seems clear that PR dialect couldn't be an extension of current RIF Core

<sandro> Harold, you're just saying you think the current Core should be *named* something other than "RIF Horn Dialect"? Or you want it to BE something different?

csma: have to go down to find common core, from which both HR dialect and PR dialect come

Harold: anyway, we have gone beyond Horn, with equality, etc.

<IgorMozetic> what about: RIF LP dialect ?

<sandro> Harold: "RIF Logic Dialect"

chris: Harold isn't saying to extend the core, merely to change the name
... since anyway, what's in core now is beyond Horn

<sandro> Michael: "RIF Logical Core Dialect"

michael: that's why it makes sense to call it logic

csma: we don't want to forego idea of having one common core

Michael: makes sense to develop a core for logic dialects, and a core for production rule dialects, both of which come from a simpler common core

<DaveReynolds> Perhaps use "base" - "PR Base" and "LP Base"

csma: we don't want the term core except for in the one core from which all PR and HR dialects spring forth.

<ChrisW> too close to "data base"

<sandro> you beat me, Dave, I was just going to suggest "base" :-)

+1 Dave

<sandro> "RIF Logic Base Dialect"

csma: base too close to database

<sandro> "RIF Logic Basis"

<sandro> "RIF Basic Logic Dialect"

<Harold> PROPOSED: To better understand what RIF Core could be, create two task forces in RIF, one focusing on a Logic Basis and one focusing on a PR Basis.

<ChrisW> "RIF Basic Logic Dialect"

<ChrisW> "RIF Basic PR Dialect"

<ChrisW> ?

csma: to Michael --- you're not proposing to extend the basis logic dialect? PR will almost certainly have notion of negation.

<DaveReynolds> ChrisW: +1 (though if we only worry about the names we are home and dry :-))

Michael: No

csma: except for name, any objection to creating these 2 task forces?
... i.e., any objection to these 3 resolutions?
... okay, hearing no objection, we'll pass the resolutions once we decide on the names.

<AllenGinsberg> 0) RIF COMMON CORE 1) RIF LOGICAL DIALECTS CORE 2) RIF PR DIALECTS CORE

csma: we'll probably just use the names basic logic and basic pr in conversation.

csma: (to allen:) basic or basis better than term "core"

<ChrisW> DaveR: naming discussions are usually longer than technical ones

<ChrisW> ...because everyone understands the discussion

<DaveReynolds> ChrisW: agreed :-)

Allen: we should have a task force for the common core, at the same time as the other two task forces, otherwise it might seem as if we're abandoning idea of common core.

<sandro> (I like Allen's suggestions, too.)

csma: we'd keep the plenary telecon

<Harold> Allen, Yes, we can have a 'twin core', and later (re)discover the 'common nucleus' underneath.

csma: but some telecons would be more relevant to logic task force; some more relevant to production rules task force.
... idea is not for two task forces to separate.
... therefore, we don't want to have a core task force: that is the working group itself.

The proposed resolutions now become:

<sandro> csma: RIF Basic Logic Dialect, RIF Basic PR Dialect

<csma> PROPOSED: To better understand what RIF Core could be, create two

<csma> task forces in RIF, one focusing on Horn and one focusing on production

<csma> rules

<csma> PROPOSED: To better understand what RIF Core could be, create two

<csma> task forces in RIF, one focusing on Horn and one focusing on production

<csma> rules

<csma> PROPOSED: To better understand what RIF Core could be, create two

<csma> task forces in RIF, one focusing on

<csma> a logical dialect and the other one focusing on production rules

<csma> dialect

<sandro> PROPOSED: To better understand what RIF Core could be, create two task forces in RIF, one focusing on a logical dialect and the other one focusing on a production rules dialect

no objection so

<sandro> RESOLVED: To better understand what RIF Core could be, create two task forces in RIF, one focusing on a logical dialect and the other one focusing on a production rules dialect

Onto second resolution

<csma> PROPOSED: Rename the current "RIF Core" draft: "RIF basic logic dialect"

RESOLUTION: Rename the current "RIF Core" draft: "RIF basic logic dialect"

Onto third resolution

<csma> PROPOSED: Create a "RIF basic PR dialect"

RESOLUTION: Create a "RIF basic PR dialect"

csma: Regarding organization of these task forces, we'll continue as we have, but will have new email topics, and new sets of actions, so we can focus discussion on one or the other.
... no need to go further, no need for formal organization
... I've been working on strawman for basic PR dialect; hope to publish it before the end of the week, to start off the discussion.
... sandro put out a strawman for this in email

4.2 ASN --> XML

sandro: In Innsbruck, we talked about requirements for RIF XML serialization. Not many requirements.
... One firm requirement: (Leora to Sandro: Missed it --- what was this requirement?)
... could pick a language that was a subset of rdf/xml,. allowing use of rdf tools.
... Harold pointed out one flaw in the way rdf does data types
... good to aim for both communities

csma: didn't see relationship between xml syntax strawman and rif serizalization strawman sent out yesterday

sandro: mapping from asn to xml schema
... and mapping from the instance level
... those 2 mappings are strongly connected, inform each other,
... but don't subsume one another

csma: in object-oriented language like ilog, and many java pr engine, we would have a rif object model conforming with asn syntax, systematically derived from xml syntax. The mapping would not go from internal model for rules in engine to xml, but from internal representation in engine to abstract asn representation to xml document.
... in that view, I don't think we need rif syntax data structure that Sandro proposed
... I would expect that: xml syntax for rif dialect is concrete syntax for that rif dialect.
... asn is abstract syntax for that same dialect.
... so I understand the first xml syntax strawman because I understand how to derive xml syntax from asn abstract syntax
... where asn syntax has classes like rules, if-then statements, conditions,
... but now, in the serizalization strawman, I see yet another syntax for rif, with none of these abstract concepts.

sandro: let's forget about the serialization strawman for now, then.
... I can go back to the way I was talking to it before.

michael: similar question to csma's
... what' s the advantage of using the rdf syntax? Just that you can use rdf tools?
... that's kind of vague ... what problem are we trying to solve?
... and what are the advantage of your solution?
... I'm having difficulty understanding what it's all about?

csma: doesn't undersatnd michael's question

sandro: Michael thinks that in the core doc, there's already a start at xml syntax

sandro: there isn't an existing syntax

michael: yes, there's something

sandro: but it doesn't give datatypes
... not in the examples

michael: so, yes, we should do it better

sandro: so that's what I'm trying to do, to nail all that stuff down, and use what another working group has spent years working on.

<sandro> stuff like "$49" and "LeRif"

csma: in the working draft, there are proposals for xml syntax, but they haven't been furthered.
... question is not just: what is xml syntax for the current draft, but how do we derive the xml syntax.

<GaryHallmark> the problem is we need some simple rules to transform a syntax in asn to a syntax in xml schema

<Harold> We already had discussed solutions for "$49" and "LeRif" here: http://www.w3.org/2005/rules/wg/wiki/A.1.1_Basis%3A_Positive_Conditions_over_Bipartitioned_Constants

csma: regarding why xml/rdf rather than not-rdf xml: if it gives us the possibility to use both kinds of tools, so that both rdf people and xml people can stay in their worlds, what's the drawback?

<GaryHallmark> the question is whether rdf/xml helps or hinders the solution to the problem

sandro: up to members of the working group to advocate for the tools they want to use.

<GaryHallmark> JAXB

< GaryHallmark> acronym for Java API for Xml Binding

sandro: for example, Gary has advocated for JAXB.

<GaryHallmark> pretty much any legal xml schema will "work" for jaxb, but it could yield a very cumbersome api

<csma> http://lists.w3.org/Archives/Public/public-rif-wg/2007Jul/0058.html

Action on Gary to check whether there would be problems with JAXB if the xml syntax would be derived from the asn in the email posted above.

csma: Gary, problem hidden in the translator?

<Harold> The translator would also need to be maintained, as we develop RIF through its lifecyle.

Gary: some confusion about how the translation would work .. .. don't know enough now to produce the translation ...

csma: Sandro, can you clarify this for Gary?

<sandro> ACTION: Sandro to produce XML Schema following his Serialization Strawman proposal -- due Julu 27 [recorded in http://www.w3.org/2007/07/17-rif-minutes.html#action01]

<rifbot> Created ACTION-328 - Produce XML Schema following his Serialization Strawman proposal -- due Julu 27 [on Sandro Hawke - due 2007-07-24].

<ChrisW> I have to leave the call.

<Harold> I noticed asn07 now being often mentioned in the discussion instead of asn06, although at F2F6 I heard asn07 was still purely experimental.

csma: there were a number of side discussions ...
... (you (Sandro?)) proposed that the root of the document should be one thing, Harold, that it should be another

Sandro: still useful to try to have this in RDF, but I know Harold doesn't agree.

<sandro> rdf:RDF to rif:RIF

Harold: it should be mappable to RDF

csma: we need to find a use case where the fact that it is rdf xml and not pure xml is a problem.

Dave: for the issue of how to extend on top of RDF: in the OWL world, the issue is not one of syntax, it's that OWL was trying to be a semantic extension of RDF --- doing that sort of extension of RDF caused semantic problems.
... But the proposal of RIF is not to do that. It's not that RIf doc would be in RDF; it's simply to use RDF to encode things .. therefore the semantic issues that the OWL community faced would not arise.

<sandro> Dave: We'd need to be very careful in getting advice from OWL community here, since they have semantic issues (with being an RDF extension) which don't seem to apply to us.

csma: (responding to Harold): if we use rdf/xml only as syntax, as a carrier, what's the benefit to the rdf people?

<Harold> RDF syntax would allow that a rule has 'extra parts', which are just not mentioned here.

dave: as sandro said, we have a well-defined self-describing syntax we can just get off the shelf.
... second, we can use various tools.
... third, it's important that metadata be interpretable as rdf and if you already have your syntactic encoding in rdf, you have that.

<Harold> ... but RIF syntax should comply to a 'closed-property assumption'.

<DaveReynolds> Harold: sure, if you take a RIF document and add additional RDF assertions then it would either be no longer a legal RIF document or the additions would be treated as metadata

csma: to harold and/or sandro:in proposed serialization, for the examples in current draft, would serizalization give different result from proposed syntax in draft?

sandro: yes, for some of the leaves, like $49. ... for information already there, it shouldn't be significantly different.

sandro: we can show examples, so people can see this better.

Harold: DaveR, re: metadata, you had shown in your 'Worked example' how to embed a pure XML syntax into RDF.

Dave Reynolds:> Harold: yes, my original proposal was to avoid the RDF/XML discussion for the bulk of the condition language but retain it at the top level to permit metadata

DaveReynolds:> Leora - sorry I was trying to respond to Harold's IRC discussion .. I can see the abmiguity!

AOB

AOB: None

next agendum.

<JeffP> +1

<sandro> +1 adjourn

Summary of Action Items

[NEW] ACTION: Sandro to produce XML Schema following his Serialization Strawman proposal -- due Julu 27 [recorded in http://www.w3.org/2007/07/17-rif-minutes.html#action01]
 
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.128 (CVS log)
$Date: 2007/07/17 16:32:46 $