RIF Telecon 29 April 2008

29 April 2008


See also: IRC log


Adrian Paschke, Alex Kozlenkov, Axel Polleres, Chris Welty (ChrisW), Christian de Sainte Marie (csma), Dave Reynolds, David Hirtle, Gary Hallmark, Harold Boley, Hassan Ait-Kaci, Jos de Bruijn, Mark Proctor, Michael Kifer, Sandro Hawke, Stella Mitchell
Igor Mozetic, Mohamed Zergaoui, Paul Vincent
Christian de Sainte Marie
Stella Mitchell


<Harold> We should talk more, write less :-)

<csma> Yes, but writing keeps the group inside the loop. I tried to propose solutions from various angles, but got little reaction from the group :-(

<Harold> Right, people can hardly keep up reading...

<csma> Well, participating to a WG is supposed to take around 20% of your time, at least (only participating, not editing or what)

<csma> On the other hand, if people do not care more, it may mean that we can just keep the solution you propose and move on :-)

<Harold> I know. For the majority, however, emails often seem to come in huge 'clusters', all in 3-5 hours, then silence again --> lack of continuity.

<Harold> Yes, let's move on.

<csma> what can we do about it? (I mean, the email flares)

<Harold> Let's plan (for the last 4 weeks or so).

<csma> Harold, I still have another proposal in my sleeve, for the case people really care about metadata, not document structure...

<Harold> Yes, you should have become a magician (maybe you are one :-)

<csma> anywhere you have object elements inside a role element, allow that object (or series of object) to be wrapped in a Block element, that would have two roles: <meta> and <wraps>, where the original object element(s) would go into the wraps element

<csma> this is based on discussion you had last week in the telecon

<csma> would impact the XML only, not anything else (how to render the metadata would be left to implementations)

<csma> what would you think?

<Harold> So far I tried to have everything from the XML also in the PS.


<csma> PROPOSED: accept minutes of telecon April 22

<csma> http://lists.w3.org/Archives/Public/public-rif-wg/2008Apr/att-0145/2008-04-22-rif-minutes.htm

Christian: Any objections to accepting minutes of last week's telecon? ... no objections

<csma> RESOLVED: accept minutes of telecon April 22

Christian: Any agenda ammendments? MichaelK said in email that he found a major bug in BLD, we will discuss that today too.


Christian: Everyone should answer the F2F questionnaire (link below)

<csma> http://www.w3.org/2002/09/wbs/38457/f2f10/

<csma> Axel, any update on F2F10?

AxelP: The weather is nice here, questionnaires should be filled out by May 10, and let me know if you need help with travel arrangements

Action review

Sandro: Action-470 continued

Christian: Action-450 pending discussion
...Action-454 continued
...Action-446 continued
...Action-452 continued
...Action-439 continued
...Action-435 continued
...Action-434 continued
...Action-152 continued


Christian: Any updates?

<AdrianP> no

DTB status

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

Christian: Axel, update on DTB?

<AxelPolleres> http://lists.w3.org/Archives/Public/public-rif-wg/2008Apr/0194.html

Axel: The link above is to an email I just sent about the current DTB status. In the email I describe three issues, along with options and proposed resolutions. I would like the WG to agree on resolutions for these issues before I continue work on the document. Once we settle on these three issues, I will be able to finalize DTB fairly quickly.

Christian: Axel, will you be here next week? We can discuss the DTB and these issues next week

AxelP: Yes, will be here

<ChrisW> ACTION: csma to put dtb on next weeks agenda [recorded in http://www.w3.org/2008/04/29-rif-minutes.html#action01]

<trackbot-ng> Created ACTION-471 - Put dtb on next weeks agenda [on Christian de Sainte Marie - due 2008-05-06].

Bug in BLD

<Harold> Bug in BLD:

<Harold> http://www.w3.org/2005/rules/wiki/BLD#Formulas

Christian: MichaelK, can you describe the major bug that you found in BLD?

MichaelK: The problem is that the syntax is not defining rules, but is defining more than rules

<Harold> "Rule implication: If f is a well-formed atomic formula and ? is a RIF-BLD condition then f :- ? is a well-formed formula"

Christian: The EBNF looks fine to me

ChrisW: Where exactly is the error?

Harold: This is in section 2.4 in the syntax; the problem is that the entire implies is again a formula. The error is in the direct specification of BLD syntax, not in the derived specification; and it is in the Mathematical English specification of the syntax, the EBNF is correct.

<josb> it is not really clear where the definition of well-formed formula ends

<josb> there should be two definitions: one of RIF-BLD condition and one of RIF--BLD rule

<Hassan> It looks like this error is a mild one ... and can be easily fixed

Christian: So, is the point that we need more reviews of the document?

MichaelK: This error has been there for awhile, and all reviewers missed it

<josb> This error was not there in February.

<Harold> Jos, maybe the error was introduced as we changed what '(uni)term' and 'formula' means after being asked to abandon visible Uniterms.

Hassan: I think the editors have the burden to show that the direct and derived specifications are equivalent

<josb> 7

<Hassan> This is undecidable ...

MichaelK: My point is that is is hard to check this document, so we need to focus now on making sure it is correct and not debate new items too much

<Hassan> On the being pressed with time, I agree that we should stay focused and give time limits for issues to be settled one way or the other

Sandro: It is OK to fix bugs such as this after last call, so we don't have to be overly concerned with finding every last one now.

Christian: Yes, bugs are inevitable, and last call will help us find them when a larger audience reviews and implements

MichaelK: Still concerned about having flaws in the design

ISSUE-51 (meta data)

<csma> http://www.w3.org/2005/rules/wg/track/issues/51

ChrisW: I've gone though the various proposals for meta data - Jos originally made a proposal, Harold and Michael modified this when preparing BLD and FLD. Since then, I, Christian, and Harold have made proposals to address the various concerns people raised. (summarized main characteristics of all the proposals)

<MichaelKifer> Harold's proposal *does* address individual rules concern!

ChrisW: The main objection to the current specification is that some people would like to be able to have a tag on an individual rule without wrapping the rule in a group

<ChrisWelty> Document ::= 'Document' '(' IRIMETA? DIRECTIVE* Group? ')'

<ChrisWelty> Group ::= 'Group' IRIMETA? '(' (RULE | Group)* ')'

<ChrisWelty> Ruleset ::= 'Ruleset( ' absolute-IRI? Metadata* Rule* ' ) '

<ChrisWelty> Rule ::= 'Rule( ' absolute-IRI? Metadata* RULE ' ) '

<ChrisWelty> CW:

<ChrisWelty> Document ::= 'Document' '(' IRIMETA? DIRECTIVE* (Group | Rule)* ')'

<ChrisWelty> Group ::= 'Group' IRIMETA? '(' (RULE | Group)* ')'

<ChrisWelty> Rule ::= 'Rule' IRIMETA? '(' RULE ')'

(Chris lost IRC connection)

<Harold> My earlier proposal (Chris and Michael just referred to): http://lists.w3.org/Archives/Public/public-rif-wg/2008Apr/0151.html

<Harold> Add:

<Harold> 'Forall' IRIMETA? Var* '(' CLAUSE ')'

<Harold> 'Exists' IRIMETA? Var* '(' FORMULA ')'

<Harold> My newer proposal: http://lists.w3.org/Archives/Public/public-rif-wg/2008Apr/0193.html

<Harold> Document ::= 'Document' '(' DIRECTIVE* Group? ')'

<Harold> Group ::= 'Group' IRIMETA? '(' (RULE | Group)* ')'

ChrisW: I would change my proposal slightly based on Christian's, but I think the syntax I proposed addresses everyone's concerns. I didn't see any feedback on it. Does this meet everyone's requirements regarding how to attach metadata (not on what type of metadata)?

JosB: I'm fine with Chris's proposal

MichaelK: My concern is the complexity of the syntax, I think it makes the syntax harder to define
... ...and also it has to be compatible with BLD and so "rule" is an inappropriate tag name

Christian: Are you saying any tag in FLD would have to be in BLD as well?

MichaelK: Yes

Harold: I agree with MichaelK's concerns. I think we should take a minimilistic approach to metadata, so that we can quickly move on to the next topics

<AdrianP> +1 for Harold

<AdrianP> given that we are running out of time and meta data is not on the critical path of BLD I agree with Harold and favour a minimalistic approach

Harold: And this issue is not specific to RIF, so we can have a metadata handle and solicit feedback from other groups, as I mentioned in my email (link below)

<Harold> "small metadata handle" proposal: http://lists.w3.org/Archives/Public/public-rif-wg/2008Apr/0193.html

Christian: Would the rule tag be needed for any other purpose than attaching metadata?

JosB: Yes, it is needed to give the rule an identifier.
... I am fine with renaming the tag to address FLD compatibility

ChrisW: Responding to MichaelK: I think my proposal does extend to BLD because it sets "rule" apart, and "group" is still there. It doesn't require anyone to use rule, or group, people can use whichever they need for their situation. And, it's a simple syntax change: it's an alternate tag name for one unit of the syntax

MichaelK: If we need a tag with cardinality 1 for rules, then why don't we need it for other constructs, such as constraints and queries?

ChrisW: Maybe we do, but I'm saying we need it for rule for now.

<josb> indeed!

MichaelK: But then we should handle constraints and queries, etc in the same way, and have a wrapper for them also

Christian: If a future dialect wants to identify e.g. queries individually, can't that can be added in that dialect?

Christian: I suggest that we indicate in the metadata whether a formula is a rule, fact query, constraint etc. How about that?

<Harold> +1 to Christian (it seems to be like http://lists.w3.org/Archives/Public/public-rif-wg/2008Apr/0193.html)

MichaelK: Yes, that seems like a good solution to me

JosB: This goes beyond BLD. My concern was identifiying individual rules.

<Harold> Yes, Christian, this is the minimalistic approach which moves us forward.

<csma> <Group><meta>identifier[content->"Rule"; otherKey->valeu]</>RULE</>

JosB: So, for BLD you would indicate whether it is a rule or a ruleset in the metadata?

<MichaelKifer> id-of-group[type->rule author->Jos]

<josb> <Group><meta>identifier[content->"Ruleset"; otherKey->valeu]</>RULE</>

<Harold> Compelling idea.

<Hassan> OK to call it meta - but why limit it to only Rule/Ruleset ... Allow it everywhere...

Christian: In summary, we would just have the one tag (name to be decided), and we would indicate type of construct in the metadata

<sandro> +1 Hassan -- okay to call it Meta, but allow it everywhere

<Harold> This way, we can distinguish not only Rules and Rulesets without proliferation of element names.

<AdrianP> it's a general proposal which allows easy extension

<ChrisWelty> Document ::= 'Document' '(' IRIMETA? DIRECTIVE* (Meta | RULE)* ')'

<ChrisWelty> Meta ::= 'Meta' META-TYPE IRIMETA? '(' (RULE | Meta)* ')'

JosB: I can live with this proposal, but I find it ugly

<Hassan> I agree with Jos - this is a hack

MichaelK: I think the proposal is elegant

ChrisW: I think Jos is saying that when you write a rule, it's going to be ugly syntactially

<josb> 'Meta' IRIMETA? '(' (RULE | Meta)*

<ChrisWelty> Document ::= 'Document' '(' IRIMETA? DIRECTIVE* (Meta | RULE)* ')'

<ChrisWelty> Meta ::= 'Meta' IRIMETA? '(' (RULE | Meta)* ')'

MichaelK: How will we derive the XML from the presentation syntax?

<ChrisWelty> STRAW POLL: this bnf plus specific metadata for Rule & Ruleset

<GaryHallmark> can't call it 'meta' because it contains rules, etc. 'meta' implies it contains only metadata

<AdrianP> yes, Garry is right

<AdrianP> Group better captures the intented semantics of this construct

<AdrianP> alternatively the type "rule | ruleset" could become an optional attribute with a defaul value "rule"

<ChrisWelty> Document ::= 'Document' '(' IRIMETA? DIRECTIVE* (Meta | RULE)* ')'

ChrisW: We are voting on the design and a minimal set of meta data, and we can add to the set later

<ChrisWelty> Meta ::= 'Meta' IRIMETA '(' (RULE | Meta)* ')'

<csma> +1 to Gary; I would rather call something neutral, like FormulaWrapper or Wrapper, or, well, Group...

<GaryHallmark> I humbly suggest calling it 'Item'

ChrisW: Let's vote on the basic design. Tag names will still be up for discussion, we don't have to settle those in this vote.
... we should allow for items that don't have metadata (e.g. facts)
... we don't want the syntax to burden people with the requirement to always use the metadata wrapper

<Harold> Document ::= 'Document' '(' DIRECTIVE* (Meta | RULE)* ')'

<Harold> Meta ::= 'Meta' IRIMETA? '(' (RULE | Meta)* ')'

<Harold> Document ::= 'Document' '(' DIRECTIVE* Meta ')'

<sandro> NONBINDING-PROPOSED: have only one wrapper construct

<GaryHallmark> what about a document of ground facts?

<GaryHallmark> is Document a wrapper?

<sandro> No, Document is not a wrapper

<sandro> NONBINDING-PROPOSED: have only one wrapper construct (by some as-yet-undetermined name), with a minimal set of metadata

<sandro> NONBINDING-PROPOSED: have only one wrapper construct [inside Document] (by some as-yet-undetermined name), with a minimal set of metadata

Sandro: What do we mean by minimum set of metadata?

<DaveReynolds> e.g. foo[rdf:type->rif:Rule] ?

Christian: For example, if we keep frame syntax for metadata, one standard slot name would be "content" and there would be standard values for that slot, such as "rule" and "ruleset"

Sandro: ok, I understand Dave's example

<GaryHallmark> so you could nest a ruleset inside a rule and only by looking at the metadata could you tell that is what is going on...

<sandro> NONBINDING-PROPOSED: have only one wrapper construct [inside Document] (by some as-yet-undetermined name), with some minimal fixed metadata, eg type (might be "Rule", "Ruleset", etc)

<DaveReynolds> -0

<sandro> NONBINDING-PROPOSED: have only one wrapper construct [inside Document] (by some as-yet-undetermined name), with some minimal fixed metadata, minimally type (minimally including values"Rule", "Ruleset", etc)

<josb> 0

<GaryHallmark> -1 I think the difference between rule and ruleset is too important to hide in metadata

<AlexKozlenkov> Seconded Gary

<MichaelKifer> The whole issue is about METADATA! What does it have to do with "hiding" rules?

<AdrianP> I would prefer that meta data is optional

<AdrianP> Group is not meta data

<AdrianP> Group might carry an attribute which defines the type of the Group content

<AxelPolleres> +0.27 (actualy being happy to get the issue from the table, it stops us progressing on other stuff for several telecons now already)

<DaveReynolds> -0 (syntactic form and the type metadata interact, it the metadata says it is a fact but looks like a rule syntactically ...)

<MichaelKifer> +1

<Harold> Syntax: Document ::= 'Document' '(' DIRECTIVE* Meta ')' and Meta ::= 'Meta' IRIMETA? '(' (RULE | Meta)* ')' , where the name 'Meta' still to be found ('Group', 'Block', 'Unit', ...)

<Harold> +1

<Hassan> ???

<ChrisWelty> 0^10^10^10^10^10

<AxelPolleres> chrisW is this a symspace?

<ChrisWelty> those are exponents

JosB: The current discussion is not only about metadata but also about rule identification

<Hassan> +1 with Jos

<Hassan> MK means that his +1 following what I wrote on the IRC does not mean he agreed with my entry above it

Christian: Jos, what is your opinion on GaryHallmark's comment?

JosB: I agree with Gary's comment

<AdrianP> +1 (for only one wrapper construct for attaching meta data)

<markproctor> +1

Christian: The straw poll did not indicate that we are close to resolution. Are there any other comments on this topic?

<AxelPolleres> I think at some point we have to put the proposals on the table and make a decision by strict majority votes.

<GaryHallmark> rolling back the Groups might help...

Harold: Maybe we should not include metadata for last call?

ChrisW: That's not the point, there are other concerns and removing metadata will not resolve them

<Zakim> sandro, you wanted to make a process comment

Sandro: The charter says we must include a way to include metadata; so we *have* to have it for last call, and it should not be that difficult.

<AxelPolleres> we don't get further with discussing this over and over.

ChrisW: Some people simply want to be able to distinguish an individual rule statement

<Hassan> I agree with ChrisW ...

<AxelPolleres> need to go, sorry.

Issue-29 (profiles in core)

Christian: There is an appendix in BLD that talks about profiles, subdialects of FLD

ChrisW: Can we move this to FLD?

<csma> http://www.w3.org/2005/rules/wiki/BLD#Appendix:_Subdialects_of_RIF-BLD

MichaelK: We can remove it from BLD, but I don't think it is related to FLD

ChrisW: It is describing things you can do with the framework (FLD)

Harold: Yes, it could be a prominent part of FLD

Christian: Core should be a separate document anyway

<Harold> Michael, the last sentence of the Overview is:

<Harold> The first of such dialects, RIF Basic Logic Dialect, is described in the document RIF-BLD.

<Harold> We can discuss the subdialects here.

ChrisW: Let's remove the appendix from BLD and optionally add more about that topic to FLD


<ChrisWelty> ACTION: remove "subdialects of BLD" from BLD [recorded in http://www.w3.org/2008/04/29-rif-minutes.html#action02]

<trackbot-ng> Sorry, couldn't find user - remove

DaveR: I don't think this resolves issue-29

ChrisW: It doesn't, but it takes it off critical path for BLD


Christian: Propose to adjourn

<Hassan> +1

<josb> +1

<AdrianP> bye

Summary of Action Items

[NEW] ACTION: csma to put dtb on next weeks agenda [recorded in http://www.w3.org/2008/04/29-rif-minutes.html#action01]
[NEW] ACTION: remove "subdialects of BLD" from BLD [recorded in http://www.w3.org/2008/04/29-rif-minutes.html#action02]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.133 (CVS log)
$Date: 2008/04/29 16:29:55 $