RDF Data Shapes Working Group Teleconference

07 Jan 2016


See also: IRC log


Arnaud, pfps, ericP, dimitris, kcoyle, labra, TallTed, hknublau, aryman


<aryman> i hear it too

* me too

Arnaud: Wellcome back everybody...let's resume our work


Arnaud: we had a good meeting at the end of december
... minutes that need to be approved

<Arnaud> PROPOSED: Approve minutes of the 10 December Telecon: http://www.w3.org/2015/12/10-shapes-minutes.html

<pfps> the minutes all looked OK to me

<pfps> +1

<aryman> +1


RESOLUTION: Approve minutes of the 10 December Telecon: http://www.w3.org/2015/12/10-shapes-minutes.html

<Arnaud> PROPOSED: Approve minutes of the 15-17 December F2F: http://www.w3.org/2015/12/15-shapes-minutes.html http://www.w3.org/2015/12/16-shapes-minutes.html http://www.w3.org/2015/12/17-shapes-minutes.html

<pfps> +1

<ericP> +1


<aryman> +1

pfps: there was a issue 115

RESOLUTION: Approve minutes of the 15-17 December F2F: http://www.w3.org/2015/12/15-shapes-minutes.html http://www.w3.org/2015/12/16-shapes-minutes.html http://www.w3.org/2015/12/17-shapes-minutes.html

arnaud: we can talk later about this

<pfps> the description of issue 115 is rather vague

next week we will have our next telecon

next F2F will be somewhere in march

arnaud: for now there is one set of dates that seem to be working
... ask people to go ahead and enter the response into the poll
... before discussing issues

Tracking of Actions

<pfps> I pushed out the due date for my action (i.e., no action on my action yet).

UNKNOWN_SPEAKER: ask if there are some actions that could inform about their status

aryman: we had a little bit of discussion about the metamodel

<hknublau> I frankly didn't do any Shapes work during the break. I expect to get back into that swing next week. I arrived back home yesterday.

kcoyle: I have later changes to use cases and requirements
... Simon and I could look at it and talk about it next week

<pfps> a notice to the WG that the document is ready for review would help

arnaud: let's take a look at shapeClass issue

ISSUE-23: ShapeClass

arnaud: everybody but Holger voted to get rid of shapeClass
... scopeClass could be improved but the group disagreed with it
... there were othre people that weren't participating at the meeting, like Ted, that wanted to review it
... the proposal is to drop ShapeClass

TallTed: I have reviewed the thread about this and what I see from Holger is a frustration
... not really agreeing with what is going on
... Shacl is not to model the world...it is a way to model data
... I want a proper phylosophical proof about that
... its nonsense to say that a shape description is not a model
... using a proper definition of fully formed ontology as a shape definition is valid use
... we reuse things all the time
... and we reuse them in context...in this instance of my data server it is...
... something in my domain
... context matters...shacl is a modelling language
... when we say that this shape is something, then it is something in a model
... I am not inclined to vote in favor of this resolution

arnaud: I understand but we don't want to go into that phylosophical discussion
... we need to move forward one way or the other
... what we have done is that when a shape is a class, the we can use the scopeClass
... there can be a point to itself
... I think you did valid points but I don't know if anyone will convince anyone at that level

<Arnaud> PROPOSED: Close ISSUE-23, dropping ShapeClass

<pfps> +1

<ericP> +1

<aryman> +1


<TallTed> -0.9

<Dimitris> +1 but woudn't mind a softer alternative

<kcoyle> +1

<hknublau> I could live with dropping ShapeClass, but as written the inference rule to infer sh:scopeClass is still in.

<Arnaud> PROPOSED: Close ISSUE-23, dropping ShapeClass, with no inferencing of sh:scopeClass

<ericP> +1

<hknublau> -1

<aryman> +1


<TallTed> -0.9

<pfps> +1

<kcoyle> +1

<Arnaud> PROPOSED: drop ShapeClass

<hknublau> -0.9

<ericP> +1

<aryman> +1


<TallTed> -0.9

<kcoyle> +1

<pfps> +1

<Dimitris> +1 but woudn't mind a softer alternative

RESOLUTION: drop ShapeClass

arnaud: now let's attack the second part

<Arnaud> STRAWPOLL: a) allow inferencing of sh:scopeClass, b) no inferencing of sh:scopeClass

aryman: can we have a recap of the inference that is being proposed?
... to what degree is inferencing done?

<pfps> making the inference is different from allowing it to happen

aryman: is it just a explicit type link or should there be other indirect relations

Holger: I believe it has the be in the shapes graph
... it needs to make possible that the shape is also a class...so subclass of rdf:class
... there is nothing that we have to worry about...it could have been done before

<Arnaud> https://www.w3.org/2014/data-shapes/wiki/Proposals#ISSUE-23:_Shapes_as_Classes

pfps: now we have just come out with a new kind of requirement
... the type must be explicit...

it is different from every thing else in SHACL

Holger: I don't see that difference

pfps: you said that it must be explicit

Holger: any subclass can also be

<aryman> here is the current spec text for inferencing: http://w3c.github.io/data-shapes/shacl/#shacl-rdfs

pfps: that's yet another different thing
... now you are going to the previous shacl

<aryman> we should use just one definition in all cases where we do "inferencing"

pfps: but now we are using a weird kind of rdfs modeling that's different from rdfs modelling

holger: in other places we have done that
... if the inference is done, that's good, but that's not our business

pfps: this really increases our need to have rdfs reasoning

<aryman> we should be consistent with sh:class : http://w3c.github.io/data-shapes/shacl/#AbstractClassPropertyConstraint

<pfps> If SHACL depends on whether a node is an RDFS class then it should be doing RDFS reasoning to determine that

arnaud: Holger, it sounds like you need to take this offline and come with a proposal about what kind of inferencing should be done

aryman: I agree with what Peter said
... if we do any kinf of class inferencing we should do it one way
... if holger is happy to use that definition of inferencing we could just use that
... if people want to do more inferencing then they should do that in another stage

Ted: a change could be to substitute sh:Class for rdfs:Class

aryman: No, they are different

Ted: sh:class has the inferencing that Holger is describing
... and that accomplish what is desired

<pfps> sh:class is a property rdfs:Class is a class, substituting one for the other does not make sense

<TallTed> { ex:MyShapeAndClass a sh:Class, sh:Shape } would imply { ex:MyShapeAndClass sh:scopeClass ex:MyShapeAndClass }

aryman: if we limit the inferencing to what we already agreed then some of the objections could be solved

<pfps> sh:Class is not part of SHACL

** Dimitris...didn't get what you said

<TallTed> @pfps - sh:class. case sensitivity is SpEcIaL.

arnaud: there is rule for some agreement...
... we could leave it as is for today
... we can use the proposals space to work on more possibilities

aryman: it is up to holger to clarify the proposal

holger: the main reason for me to accept the discussion is that Peter has threaten to reopen the whole WG

<TallTed> actually... what we do need is sh:Class (a rdfs:Class) which requires no reasoning beyond what is required for sh:class (a rdfs:Property)

pfps: I am confused about what has been proposed
... if someone wants to propose something it must be better defined

<hknublau> OK I will start a proposal.


arnaud: where are we with the draft and the spec?

<pfps> I think that republishing soon is a good idea.

holger: as I wrote earlier I was in holydays so I regret to inform that I didn't do more work on it
... after a few days I will have more free time
... we didn't work yet about the metamodeling

arnaud: it would be nice if by next week you could inform about what has been implemented and what haven't

aryman: is there something where the resolutions are summarized?

arnaud: the compilation is in the web page

<aryman> Resolutions page: http://www.w3.org/2014/shapes-resolutions

arnaud: let's talk about a issue that aryman brougt up

ISSUE-49: scope/unscoped shapes

aryman: a shape has a scope and a filter
... and the usual constraints...the question is about the filter
... we should make the filter to be associated with the constraints, not the scope
... we first apply the filter criteria, then the shape passes

the filter could be an if-then construction

scribe: filter's should be as preconditions
... this relates to issue 49

<pfps> I don't think that the status of filters has much to do with issue 49

arnaud: Dimitris, is that the issue you raised?

Dimitris: He wants aryman to clarify what he proposed

aryman: you are given a node and apply a shape to it...you evaluate the shape to it

<pfps> A node validates against a shape if it does not satisfy the filter or does satisfy the constraints

aryman: the scoping mechanism is about computing which nodes in the graph should be checked

Dimitris: my issue was more about the scope relation with the filter

aryman: the filter is always used when you compute the scope but is also used when you apply the shape

Dimitris: my issue was about how to relate scopes with constraints and filters
... and valueShape

aryman: scoping is a way to enumerate a set of nodes that you want to evaluate
... once you get those nodes, you apply the filter

arnaud: the filter is like a pre-condition to apply the rest of the shape

aryman: if you don't differentiate between those two: scopes and filters then there is no need to have them both

<hknublau> Why no straw poll?

<Zakim> ericP, you wanted to ask about shape annotations

ericP: there are two uses of this
... we use the shapes as a typing mechanism on the nodes
... there 2 ways in which scopes can be used
... in shacl I've seem two uses of filters

arnaud: I think the proposal may not be so mature to do a straw-poll
... it seems that the interaction between filters and scopes should be clarified

Ted: I am not understanding the flow
... I would like to see an example from arthur

aryman: this is described in the spec to some degree
... but there is some ambiguity
... there is a diagram in the spec

<aryman> see the diagram above 2.1: http://w3c.github.io/data-shapes/shacl/#scopes

ISSUE-115: ClosedShape

arnaud: issue 115 about closed shape
... there is no description
... lets take to review it


ericP: the goal of closed shapes is to address the problem where one post something that is not what one expects, then it is rejected
... that has more information than the one that is described by the shape
... the point of closed shape is to have a flag that one is not interested in other data than the one that is mentioned in the schema

arnaud: part of the issue is at what level it is specified...shape?

aryman: I didn't quite get what eric said...is he saying that the current definition is wrong in the spec?

ericP: the current definition doesn't do everything that one would want
... I would like it to be defined about predicates that are defined in the schema

<TallTed> I thought the point was to say "this node must only have relations that match what I've included in my shape, and nothing else; else it fails validation"

ericP: I support your observation that it should be associated with shapes

aryman: it is a kind of constraint but from a language design point of view it should be associated with a shape
... it is mainly a syntax thing
... about where we put this property

holger: it is at the level of constraints, instead of shapes
... but both views could be valid

** I lost what Holger said

** Holger, maybe you want to write it?

<hknublau> hknublau: TO summarize, I believe all constraints that actually produce violations should be grouped with the same syntax, under sh:constraint, sh:property, sh:inverseProperty.

<hknublau> ... The case to introduce an exception for sh:closed is not strong enough IMHO.

** thanks

pfps: asks about the status of the tests

Test suite

pfps: I have a bunch of tests that I would be happy to put in the test harness

ericP: I will show you what we are using in the ShEx test harness

<ericP> https://github.com/shexSpec/shexTest/blob/master/validation/manifest.ttl

<hknublau> Isn't the latest spec at https://github.com/w3c/data-shapes/tree/gh-pages/data-shapes-test-suite ?

<ericP> https://github.com/shexSpec/shexTest/blob/master/validation/manifest.ttl#L2674

ericP: describes the structure of the tests manifest

pfps: it doesn't fit well in shacl

<Arnaud> http://w3c.github.io/data-shapes/data-shapes-test-suite/

ericP: that's the original document that we are using for the ShEx test

pfps: I will take a look at it and write some tests

<Arnaud> trackbot, end meeting

Summary of Action Items

Summary of Resolutions

  1. Approve minutes of the 10 December Telecon: http://www.w3.org/2015/12/10-shapes-minutes.html
  2. Approve minutes of the 15-17 December F2F: http://www.w3.org/2015/12/15-shapes-minutes.html http://www.w3.org/2015/12/16-shapes-minutes.html http://www.w3.org/2015/12/17-shapes-minutes.html
  3. drop ShapeClass
[End of minutes]

Minutes formatted by David Booth's scribe.perl version 1.143 (CVS log)
$Date: 2016/01/14 21:20:30 $