RIF telecon 29 January 2008 -- 29 Jan 2008

<Harold> zakim NRCC is me

<csma> Scribe: Adrian Paschke

<csma> Scribenick: AdrianP

<csma> http://lists.w3.org/Archives/Public/public-rif-wg/2008Jan/att-0103/22-rif-minutes.html

ChrisW: Last weeks minutes - any objections?
... minutes accepted

<ChrisW> PROPOSED: accept minutes of Jan 22 telecon

<ChrisW> RESOLVED: accept minutes of Jan 22 telecon

ChrisW: Other admin business? no

CSMA: Action review moved to 3rd

liason

ChrisW: Jos any news about OWL-RIF?

<AxelPolleres> ;-)

ChrisW: Action review

<josb> I am currently having problems with zakim; I will continue trying to dial-in

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

ChrisW: Action 405 in the agenda
... Action 404

Harold: Done

CSMA: still pending discussion

ChrisW: Action 403 complete
... Action 402 continued

<csma> 402 is continued

ChrisW: 401 done
... PR developers only 2 weeks left for providing rule selection strategy
... several open actions
... all other actions are continued

CSMA: Pending review

ChrisW: F2F9

csma: nothing to add on F2F9

cmsa: any questions about procedure to book hotel?

harold: will there be anything on the arrival date?

csma: I won't be at ILog that day
... nothing arranged
... maybe a presentation of ILog

harold: will Hassan and Philippe around

csma: Hassan will be around
... it is right before the French school vacations

ChrisW: other discussions
... Issue 46

<ChrisW> PROPOSED: to close issue 36 without action (that is, direct mapping between presentation and XML syntaxes, e.g. presented as a table).

ChrisW: discussion on this issue?
... any objections?

<ChrisW> RESOLVED: to close issue 36 without action (that is, direct mapping between presentation and XML syntaxes, e.g. presented as a table).

ChrisW: Issue 44

<csma> ACTION: ChristopherW to close issue 36 [recorded in http://www.w3.org/2008/01/29-rif-minutes.html#action01]

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

<csma> ACTION: CWelty to close issue 36 [recorded in http://www.w3.org/2008/01/29-rif-minutes.html#action02]

<trackbot-ng> Created ACTION-406 - Close issue 36 [on Christopher Welty - due 2008-02-05].

ChrisW: most people are tired of this issue.
... proposed resolution from last week

<ChrisW> PROPOSED: to close issue 44 by removing named argument uniterms from BLD.

ChrisW: most people are leaning simplifying things

Harold: discussion on the mailing list
... several examples
... new insights
... point is RIF is an interchange format
... not putting the burden on the translators

ChrisW: any other engines than OO jDrew

Harold: Michael is not here, but he confirmed that rel. databases are a use case

csma: I responder to Michael and made a proposal
... small burden on the translators for named arguments
... anyway you would put the burden on the translators, but not on the translators of systems without named arguments

Harold: I think you proposal should be discussed

<Harold> p[bar,foo]("abc", 2)

Harold: but there was no discussion yet

csma: proposal is that, if there is standard uniterm then there is an optional list of arguments which might be ignored
... the burden is on systems who support sloted uniterms
... systems which use slotted uniterms could rebuild from the list
... others might simply ignore it
... very small burden on the translators from slotted uniterms to RIF

Harold: Don't understand that this meta data could be ignored
... remind you on signature, ignoring the signature means loosing information

csma: could be ignored by systems which are not able to use it

<sandro> (the proposal was mine, I guess, but came up in conversation with csma)

Harold: in the eMail there seems to be a contradicting between meta data

csma: yes, it's Sandros proposal

Harold: in this proposal you can no longer distinguish from the ordered and loose information

csma: if it does not fly we can not use it

<sandro> good point, Harold --- it's a lossy transformation, because (a->x, b->y) and (x,y) appear the same to systems ignoring the slot names.

csma: then only one proposal remains

Axel: I do not object to slotted uniterms
...but they are clearly defined, so why remove them

<DaveReynolds> Sandro/Harold: do systems really use the same name for both positional and named uniterms? Surely renaming apart would be needed in such cases anyway?

Axel: I don't have a strong oppinion on that

ChrisW: you would not object
... Igor would remove its objection

Michael: Don't understand why we should remove them

<AxelPolleres> +1 to Michael, I think they are simple enough.

ChrisW: reason --> simplify BLD

Michael: But they don't have to be supported from implementors

<sandro> -1 mkifer

Michael: they don't have implement that- BLD is a container

<AxelPolleres> +1 to Michael again, I support that BLD is a container.

Michael: different vendors have to support the dialects they want

<csma> -1 to everybody not being required to implement all of BLD

<sandro> -1! BLD has to be implementable.

<Harold> Dave, BLD does not differentiate 'alphabets' of Constants used for positional vs. used for named uniterms (it's Hilog-like and uniform).

<DougL> -1 to Michael

<AxelPolleres> Was there anybody who said they would implement all of BLD?

ChrisW: your are supposed to translate a RIF rule set into your own language

<josb> -1 to Michael

<sandro> If there isn't, Axel, then lets stop working on BLD right now.

ChrisW: translate from BLD into language

Michael: you might translate from a sub dialect

<Harold> We need to define Core as *subset* of BLD.

ChrisW: there are no subdialects

<AxelPolleres> I anyway think that BLD is ready and we should go on and define core and extensibility for the remaining time.

Michael: With framework we can easily define a dialect

<AxelPolleres> so, sandro: I agree.

<Harold> At the last f2f we decided to work on a Core.

csma: you can define any dialect, but then there is interoperability

<sandro> Also -- we have a resolution to have a "handful of dialects"

Michael: suppose you remove the named arguments

<AxelPolleres> yes, but we discuss still for weeks now nitty gritty details on changing/cutting BLD.

Michael: it is a burden to translate into this cut-down BLD

<Harold> Core = BLD - Equality - Frames - Slots

csma: if there is a need for additional features we will later add this feature

<DaveReynolds> Harold - I understand that, my point was that I would imagine most system and most users would use different names for the two different cases so the times where there is an aliasing clash to resolve seem very rare and easy to handle

Michael: Why then care about equality etc.

<AxelPolleres> harold: - fuinction symbols

Doug: appearance of simplicity does not mean

<AxelPolleres> I can take over...

<AxelPolleres> scribenick: AxelPolleres

<scribe> scribe: Axel Polleres

harold, can you repeat that on the irc?

<Harold> So Doug wants to keep slotted uniterms.

<DougL> Correct

<Harold> I want to emphasize that we will need a subset, the Core, which will not have Equality and will not have slots. So we can and should keept both in BLD.

Harold: What about to propose a resolution to leave them in?

<Harold> Yes,

<ChrisW> PROPOSED: to close issue 44 by removing named argument Uniterms from BLD.

so, we had two objections against removing and one against leaving them...

so, we should ALSO ask the other way around.

<sandro> 0

<DaveReynolds> +0

<josb> +1

-0

<Harold> -1

<DougL> -1

<sandro> (+1 from Gary)

<Michael_Kifer> -1

<csma> +1

<mdean> 0

<sandro> Gary: if we have N.A.U's we really have to tell people how to interchange them with normal Uniterms. I would probably object to leaving them in.

<sandro> Gary: Implementors will go off and do what they want, without interoperation.

<josb> How come we go from 1 person objecting to 3 persons objecting?

yes, it is not clear how these are related... so ... what?

josb, I even changed my objection from -1 to -0 this time, so more people switched to leaving them in. I honestly really think we should close this issue and work on with what we have in order to not loose more time.

<DaveReynolds> +1 to Gary, a well articulated explanation of what interoperability and RIF is about

<josb> Well, there was never any agreement within the working group to include named argument terms. So, there's not really any reason to include them.

we do care because BLD makes up a clean implementable framework for logical dialects!

josb, that's why I said that we should propose a resolution the other way around, and see what happens then.

<Harold> Chris, what about Equality?

<Harold> We dont know if anyone will implement it completely?

<csma> harold, there will be a Core, without equality (but without logic fct either)

<Harold> But it's in BLD (because we know it will be out of Core).

<Harold> Same should be kept for Slotted Uniterms: they are MUCH more easier to implement.

<Harold> We can hardly keep Equality but omit Slotted Uniterms.

<sandro> AxelPolleres, I thought you were scribing.....?

<csma> actually, if we wanted to have more in BLD than Core+equality+logic fct, that would be negation, I guess, not NAU...

<DougL> +1

ChrisW: let us see how the proposal works the other way around.

<sandro> STRAW-PROPOSED: to close issue 44 by KEEPING named argument Uniterms in BLD.

MichaelK/ChrisW/sandro: some discussion before on whether BLD should be implementable fragment for all for interchange or not

<DougL> I won't be here next week, probably (trip), but please consider my proxy for it as +1

<sandro> 0

<DaveReynolds> -0

<GaryHallmark> -1

<Harold> +1

<Michael_Kifer> +1

<josb> -0.6

<mdean> 0

<Harold> (+1 from AdrianP)

ChrisW: Harold and Doug, neither of you objected last time.
... something changed?

DougL: saw some useful examples when I further thought about it.

ChrisW: any specific languages you're thinking of?

<Harold> Use Cases: CLIPS, Relational Algebra, 4 from NRC.

DougL: uncompatible evolution for languages/rulebases with large numbers of arguments.

<Harold> CLIPS is supporting slotted uniterms.

ChrisW: Is there anything except ooJDrew?

DougL: cyc? supports named args in uniterms.

Harold: I rediscovered some use cases.

<Harold> Keys (local to a table) are not OIDs (global).

GaryHallmark: All use cases typically have hidden some kind of object identity.

MichealK: the issue is not that we can map it, the issue is, what does it take to do the mapping... you need for instance function symbols.
... why then have features like frames, etc at all, all can be done with positional terms.

GaryHallwmark: my system only has frames, no uniterms.

MichealK: if we mangle features like frames and uniterms, we will loose roundtrippability.

<DougL> Philosophers continually wrangle about this issue here (Davidsonians vs. non-Davidsonians)

MichealK: but if this ok, then simply let's choose positional uniterms.

ChrisW: Let's move on.

(scribe cap off): why not just leaving it, we had two polls pointing in this direction?

OWL - RDF compatibility

ChrisW: Let's talk about RIF-OWL compatibility task force.

josb: Last time I talked about OWL full vs OWL DL compatibility.
... annotation properties are different from binary predicates, so another possibility is ignoring them

Builtins

ChrisW: let's move on to issue 40.
... What is the status of errors?

Christian: ... summarizes 3 proposals.

ChrisW: first-order or third truth-value? opinions?

josb: the issue only comes up if we talk about partial functions.
... for total functions, there is no problem.

michaelk: this is not true.
... it is not only for functions, but also for predicates, which are only partially defined.
... most systems give errors for e.g. adding strings with numbers, so if we define some different (2-valued) behavior for these predicates, we crate additional burden for implementors.

ChrisW: Christian just said that we shouldn't fix it in the model theory.

josb: I havean idea here... whenever a variable assignment is not allowed, the satisfaction is not defined.

<csma> ACTION: jos to draft a strawman on error [recorded in http://www.w3.org/2008/01/29-rif-minutes.html#action03]

<trackbot-ng> Sorry, ambiguous username (more than one match) - jos

<trackbot-ng> Try using a different identifier, such as family name or username (eg. jdebruij2, jderoo)

<csma> ACTION: jdebruij2 to draft a strawman proposal on error (in ExtTerm) [recorded in http://www.w3.org/2008/01/29-rif-minutes.html#action04]

<trackbot-ng> Created ACTION-407 - Draft a strawman proposal on error (in ExtTerm) [on Jos de Bruijn - due 2008-02-05].

ChrisW: action on jos to draft proposal.

Dave: using the predicate form rather with additional argument for the value is preferable... (?)

<ChrisW> P(?x ?y ?result) vs F(?x ?y) => ?result

<josb> indeed, equality is not required

<Michael_Kifer> p(?X, 1+2).

<Harold> Special case: Equal(?X, 1+2).

<Michael_Kifer> p(?X) :- q(?X,1+?X).

<Harold> That Equal could be defined in one fact: Equal(?X, ?X).

<Michael_Kifer> :- is(1+2,?Y)

MichaelK: I think we don't need equal for functions, but rather an assignment built-in.

<csma> +1 to michael: what Dave needs is assignment, not equality

josb: don't understand why we would need the assignment.

<DaveReynolds> p(?z) <- q(?x), r(?y), add(?x, ?y, ?z).

<Harold> Builtin fcts can (normally) not be called with any argument being free.

ChrisW: assignment is simpler, because it is simpler than full equality.

<josb> p(?x+?y) :- ...

ChrisW: and to pass over values to different variables.

josb: works equally with functions.

Dave+Jos: assignment/functions both solve that issue.

Sandro: one problem with predicates is: if bld is a proper extension of core than functions and predicates might be overlapping (?)

<Harold> I also agree, ideally no duplication of builtins as fcts and preds.

josb: preference for not having functions.

<GaryHallmark> +1 to functions

csma (without chairhat): in PRD preference for functions.

<DougL> +1 for having functions and predicates both

<DougL> exactly

josb: only preds or only function symbols it not an option.

<csma> +1

<DougL> +1

<ChrisW> having both

ChrisW: who is in favor of having both?

<Michael_Kifer> +0

<josb> -0

<ChrisW> sandro: +1 (on phone)

<DaveReynolds> -0.5

<Harold> 0

<GaryHallmark> we are talking about builtins only

<mdean> +0

<GaryHallmark> is a predicate == boolean function?

<csma> yes

<DougL> one could have a predicate sumEquals and a function Plus

we are voting on what now?

:-)

<Harold> In Core, we could have the kind of Prolog-like 'is' primitive Micheal mentioned: then functional builtins would work in Core, too.

<csma> dougl, thats exactly what i call duplicating every fct

<DougL> do both

<GaryHallmark> I vote for a + b, not a + b = c

<josb> Harold, see my example above; we do not need is in Core

<Harold> You mean by flattening?

<DaveReynolds> Jos: that does imply repeating sub expressions, whereas with variable binding you just repeat the variable

I sense that there are several questions: a) whether to generally model functions as predicates. b) whether the same built-in should be (dis)allowed to have both a function and a predicate version.

<josb> when you use 'is', the formula is just as flat as when using functions directly

ChrisW: volunteers for an action?

<josb> Dave, that is correct

I can try...

but I am very unsure whether I understood it, to be honest :-)

<csma> ACTION: axel to draft an emal separating the different issues in the question about fct vs predicate forms for builtins [recorded in http://www.w3.org/2008/01/29-rif-minutes.html#action05]

<trackbot-ng> Created ACTION-408 - Draft an emali separating the different issues in the question about fct vs predicate forms for builtins [on Axel Polleres - due 2008-02-05].

ChrisW: we will not spend much time on the named arg issue next week ,but need to turn to more urgent issues.
... adjorn.
... publication plan status?

michealk: did a lot of work on the framework recently. BLD should be short base on that.
... framweork by end of the week (sunday)
... short BLD will then be much shorter than the original.
... I am not deleting stuff, but create new documents (copy-pasting).
... hope also BLD will be in shape by end of the week.
... will publish/adapt links on the wiki.

ChrisW: keep link BLD for the new one, and have a link oldBLD for the old one.

Sandro: on wiki migration.

MichaelK: I would prefer to work on html anyway.

sandro: in mediawiki you can use a flag to use simply normal html.
... I just run through wikiTR and then put the html into mediawiki.

- DRAFT -

RIF telecon 29 January 2008

29 Jan 2008

Attendees

Contents

liason

OWL - RDF compatibility

Builtins

Summary of Action Items

Scribe.perl diagnostic output