RIF Telecon 23-Dec-2008

23 Dec 2008

See also: IRC log


Jos DeBruijn, Chris Welty, Hassan Ait-Kaci, Axel Polleres, Harold Boley, Michael Kifer
Dave Reynolds, Gary Hallmark, Adrian Paschke, Leora Morgenstern
Chris Welty
Hassan Ait-Kaci




<ChrisW> Scribe: Hassan

Negative Guards

Jos: update on the OWL document - things are going "smoothly" - work on on-going issues e.g. negative guards

Chris: asking what specific restriction on NGs we need to have

Jos: yes - some restrictions are in order for it to have tractable guards
... leaning toward having such NG's

Axel: if NG's are restricted to literal only, then they're ok.
... Both and Nega. guards need to have either T or F - nothing specified otherwise
... NG issue may become obsolete if we have LP and Neg. by failure
... would rather drop them than having this "crooked" version...

ChrisW: this WG is not designing an LP dialect

Jos: Dave wants to make the ontologies independent from the OWL/RL rule processing

Axel: Dave said that we would be ok with the limited NGs

Jos and Axel: discuss the fine differences in re. data types

ChrisW: whos feels strongly about dropping NGs altogether?

Jos: not me

<josb> isNonIntegerLiteral

<ChrisW> PROPOSED: Change all negative guards to work on the literal domain only, e.g. isNotIntegerLiteral

<ChrisW> (for next telecon)

<AxelPolleres> http://www.w3.org/2005/rules/wiki/YoungParentDiscount_1

<AxelPolleres> http://www.w3.org/2005/rules/wiki/YoungParentDiscount_2

Axel: discussing the NG uses in his examples
... these examples need to know what ages are not integers in order for it to be able to compute the age differences

Chrisw and Hak: maybe this is a bit contrived?

Axel: we need to specify what to do otherwise ...

<josb> take any of the disjunction cases

ChrisW: still doubtful - we need a really uncontroversial example where isNotInteger is needed without question

Axel: such are common examples in Data Models fron the Net

<josb> http://www.w3.org/2005/rules/wiki/Disjunctive_Information_from_Negative_Guards_1

Jos: what about the UCs involving disjunction in negative guards?

<AxelPolleres> As opposed to ChrisW's argument that the YoungParent use case was about bas data modeling... I oppose that cleaning up messy data is a perfect UC for RIF, IMO.

<josb> http://lists.w3.org/Archives/Public/public-rif-wg/2008Dec/0086.html

<AxelPolleres> One example is the use of dc:creator in RDF data which some people use with strings, others to refer to an RDF resource which is a foaf:Person with a name.

ChrisW and Jos: discussing the meaning of negated types in his examples

Axel: finds it more confusing
... must specified when it is true/false or undetermined

<josb> isNonIntegerLiteral is false for everything that is an integer or a non-literal

<josb> isNotInteger is false for everything that is an integer

Jos: objects to Axel's proposal

<AxelPolleres> I didn't make a proposal ?!?

ChrisW: I understand what Jos is saying

Axel: I wanted to clarify things between Jos's point and alternative meanings

Jos: argues for his proposal as being sound and useful

ChrisW: I now understand the Test Case and I now think it is non contrived and a nice one

Jos: I am so happy :-)

<ChrisW> PROPOSED: Change all negative guards to return true only for literals that are not of the type, false for non-literals

<AxelPolleres> I still think that the use case is contrived for the normal user.

<josb> it's a test case, not a use case :)

Axel: discusses Jos's proposed Test Case use of Negative Disjunction in guards

ChrisW: wonders what the consequences of allowing/forbidding such guards would be

Jos: cannot force an object to be an integer

MK: why do we need NGs in BLD?

ChrisW: needed for OWL

MK: I seem to remember we discussed this earlier this year though not exactly what we discussed

<josb> DaveR is the biggest proponent

ChrisW: Next: DTB/Builtins for OWL/RL
... less-than or compare?

<Michael_Kifer> may be Dave could send an email explaining the issue? Certainly OWL-RL could not have been the reason back in February

ChrisW: it was already sent by email
... what about string less-than or compare?
... make things uniform across datatypes

Jos: who wants what?
... Objection to making the language more complex

Axel: less-than, greater-than (or equal) are now available for all types that have comparisons

Jos: make such things more generic/abstract

Axel: there is a task force

ChrisW: The Abridged Syntax TF is the one

Axel, ChrisW: discussing the choices made for having some operators but not others, redundancy, etc...

<josb> +1 against

ChrisW: Who's for the string operators?

<AxelPolleres> +q

ChrisW: Who's against the string operators?

Jos: finds them unneeded and unclear
... I don't care dropping them if we may define them with others

Axel: can agree to leave them for now ?

Jos: I prefer dropping them now

Axel: I introduced them in because Gary asked for them - but I don't object dropping them

<ChrisW> ACTION: Chris to update ISSUE-67 to indicate discussion is postponed until the presentation syntax is finished [recorded in http://www.w3.org/2008/12/23-rif-minutes.html#action01]

<trackbot> Created ACTION-677 - Update ISSUE-67 to indicate discussion is postponed until the presentation syntax is finished [on Christopher Welty - due 2008-12-30].

Jos: I do not think Gary minds either way (but he's not here today)

<AxelPolleres> good, let's just propose to drop them next time and ask Gary for an opinion explicitly.

More general builtins

<AxelPolleres> "Editor's Note: It is still under discussion in the WG whether an additional predicate pred:hasNotDatatype should be added, cf. ISSUE-80."

ChrisW: Dave suggested that the OWL/RL would be considerably simpler if we had such predicates

<AxelPolleres> http://www.w3.org/2005/rules/wg/track/issues/80

Jos: such may be useful, but we should choose between two kinds of guards, though I slightly prefer what we have now

<ChrisW> isOfType(?x, ?type)

<ChrisW> isNotLiteralOfType(?x,?type)

Axel: agrees that this would simplify and give it some parametricity

<ChrisW> isLiteralNotOfType(?x,?type)

Axel: discusses elegance ...

<AxelPolleres> slight preference for onlyu having more general and get rid of guyards and neg guards as a whole.

ChrisW: I hear Jos for more generality. Axel for more specificity.

Hak: I agree with Jos

<josb> I prefer isInteger; Axel prefers isType(?x, xsd:integer)

<ChrisW> Axel: guards that take type as an arg

<ChrisW> Jos: guards that have type in the name

<AxelPolleres> my argument is for maintenance nightmare...

<AxelPolleres> ... I am for more general.

<AxelPolleres> yes!

Hak:I was wrong: in fact, I agree with Axel!

<AxelPolleres> +1 to MK

ChrisW: I am hearing support for the "isType/isNotType" generic notation

<AxelPolleres> LiteralHasDatatype and LiteralHasNotDatatype

<ChrisW> PROPOSED: add isOfType and isNotOfType (based on resolution of issue-79) and remove specific type-named guards (e.g. isInteger)


<AxelPolleres> isLiteralOfDatatype

<ChrisW> isLiteralOfType/isLiteralNotOfType

<ChrisW> isLiteralOfDatatype/isLiteralNotOfDatatype

Hak:I prefer ChrisW's

<ChrisW> isHardToType

<josb> (i) isLiteralOfType/isLiteralNotOfType

<ChrisW> +1 shorter

<josb> (ii) isLiteralOfDatatype/isLiteralNotOfDatatype

<ChrisW> +1 i

Hak:+1 for (i)

<Harold> +1 i

<josb> 0

<AxelPolleres> +1 for (ii), no objection to (i)

<ChrisW> PROPOSED: add isLiteralOfType and isLiteralNotOfType (based on resolution of issue-79) and remove specific type-named guards (e.g. isInteger)

<AxelPolleres> +1


<josb> +1

<Harold> +1

Axel: what about a resolution with negative guards?

<ChrisW> PROPOSED: add isLiteralOfType and isLiteralNotOfType (based on resolution of issue-79) and remove specific type-named guards (e.g. isInteger, isNotInteger)

<AxelPolleres> ok!

OWL-RL builtins

<AxelPolleres> does that mean I can start implementing this? :-)

Jos: What are the issues?
... Shouldn't this be datatypes?

<AxelPolleres> (BTW: we still need a RESOLVED: for the minutes, or you only want to resolve it next time?)

ChrisW: this issue then is about datatype
... what OWL datatypes do we have to support that we do not already?

Jos: Dave had reservations about these issues (implementability)

ChrisW: Boris Motik seemed to be willing to drop some of their stuff

Jos: I did not get the same feeling

<ChrisW> We are not passing resolutions today

<ChrisW> just proposing them for next telecon

Jos: the XML schema explicitly states that applications are free to interpret some of these datatypes
... also has a list of necessary things that implementations must support

ChrisW: has anyone listed the discrepancies between OWL/RL and RIF datatypes?

Jos: cites examples of such

Jos and ChrisW: review some weird OWL/RL datatypes ...

ChrisW: anyone on the call has an opinion of these?

<AxelPolleres:> that anyURI is a not subtype of strings is IMO kinda weird... isn't it?

Jos, ChrisW, Axel: discussing data typing in OWL/RL

Jos: a priori has no objection in re. data types except for the date/time data type

<AxelPolleres> ok, at least we have agreed on proposing some resolutions next time. :-)

ChrisW: any other discussion?

<AxelPolleres> MERRY CHRISTMAS!!!

Hak: Happy everything!

Summary of Action Items

[DONE] ACTION: Chris to update ISSUE-67 to indicate discussion is postponed until the presentation syntax is [recorded in http://www.w3.org/2008/12/23-rif-minutes.html#action01]
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.133 (CVS log)
$Date: 2008/12/23 17:29:26 $