See also: IRC log
<scribe> scribe: cygri
<pfps> Anything about next F2F?
<Arnaud> PROPOSED: Approve minutes of the 26 February Telecon: http://www.w3.org/2015/02/26-shapes-minutes.html
Arnaud: ArthurRyman has retired from IBM and is now an Invited Expert.
<pfps> Minutes of 26 Feb look OK
RESOLUTION: Approve minutes of the 26 February Telecon: http://www.w3.org/2015/02/26-shapes-minutes.html
subtopic: Minutes of F2F
<pfps> Minutes of F2F look good to me
<Arnaud> PROPOSED: Approve minutes of the F2F2: http://www.w3.org/2015/02/17-shapes-minutes.html http://www.w3.org/2015/02/18-shapes-minutes.html http://www.w3.org/2015/02/19-shapes-minutes.html
RESOLUTION: Approve minutes of the F2F2: http://www.w3.org/2015/02/17-shapes-minutes.html http://www.w3.org/2015/02/18-shapes-minutes.html http://www.w3.org/2015/02/19-shapes-minutes.html
subtopic: Next meeting
Arnaud: Next call will be March 12th
subtopic: Next F2F
Arnaud: We left the previous F2F with open question regarding dates and location
<pfps> I think that IBM CAS north of Toronto would be an excellent location as well
Arnaud: pfps offered options to host, including on a small island that sounded nice but maybe not ideal
... and cygri offered to host in Raleigh
ericP: I was hoping for a meeting in Europe, no more budget for international travel
Arnaud: I’d prefer to keep us in NA this time and Europe next time
... and meet in May
... suggest doing a Doodle poll
<pfps> poll sounds like the best method (but, of course, all methods are bad)
ericP: Sounds good to me
<scribe> ACTION: Arnaud to set up Doodle poll for date/location of next F2F [recorded in http://www.w3.org/2015/03/05-shapes-minutes.html#action01]
<trackbot> Created ACTION-16 - Set up doodle poll for date/location of next f2f [on Arnaud Le Hors - due 2015-03-12].
<pfps> ACTION-15 looks done to me
Arnaud: ACTION-15 is done, propose to close
<trackbot> ACTION-15 -- Richard Cyganiak to Propose a rephrasing of req 2.11.7 -- due 2015-03-05 -- CLOSED
hsolbrig: No progress on ACTION-11 and ACTION-12
<pfps> SWM isn't an official action, i think
hsolbrig: will get to that in next couple of days
Labra: I proposed a different wording for 2.11.7
<pfps> the action was to propose something, and that was done. The action was not to fix the world.
<trackbot> ISSUE-21 -- What is shape expressions? -- raised
Arnaud: We closed just cygri’s action to propose a wording. This doesn’t mean the wording or requirement is accepted.
... pfps pointed out that there are multiple definitions of ShEx
pfps: People say “ShEx” and it’s not clear what is meant
... Very different things with the same name
ericP: We have implementations that seem to give the same results on the same input. How different are they really?
pfps: They are very different. Different semantics, different intuitions, etc
ericP: Implementations based on algebraic semantic and on the W3C submission were used and gave the same results
pfps: The member submission is completely open, the algebraic one completely closed.
Labra: Which is the algebraic one? My axiomatic proposal is open. Some of my earlier papers had closed semantics, but now I am trying to stick to open semantics.
... General feeling is that we are talking about the same thing
... There may be corner cases that need to be worked out but the goal is to have the same semantics
... The goal is to have open semantics
... In my semantics, there is a construct to close a shape, but that is just an option
... Experimental option. Goal is to have same semantics as ericP.
pfps: Telling example: I have a b-filler that is c and a b-filler that is e
... In member submission, this is never satisfiable, in the other one it is
[scribe is probably getting this wrong]
ericP: Uniqueness criteria were changed. In submission we had single occurence requirement. This didn’t work, so we extended it.
... There was lots of discussion about this.
... But I have no idea what the answer to these cases would be in SHACL.
Arnaud: The status quo is we have different versions of ShEx.
... I am not sure that the details have any bearing on the discussion at this time
... When you say “ShEx”, you have to be careful not to assume too much
... Most of the differences don’t matter now
pfps: We are already talking about closure, about And, about Or, and nobody knows what they mean
ericP: Well we will have to figure them out one by one
pfps: Nobody knows what we are really talking about. Everybody thinks, “I like this”, and then later figure out that it’s something completely different
Arnaud: Until we write down all details in the spec, there is room for misunderstanding
pfps: It’s ever so much worse because there are three versions of ShEx
... If anyone talks about ShEx, link to a spec
Arnaud: That seems fair enough
ericP: Jose, does your semantics capture [???]
<Arnaud> PROPOSED: close ISSUE-21, asking for people to clarify which version of ShEx they refer to when they use that term
<pfps> That's fine by me.
Labra: I’m trying to. Work on ShEx goes all the way back to the workshop, and these are the things we are now trying to clarify
... We also have a glossary that is helpful to clarify what we mean by Open/Close, Or, etc
Arnaud: For now, how about we always reference a version?
... Until there is convergence?
RESOLUTION: close ISSUE-21, asking for people to clarify which version of ShEx they refer to when they use that term
<pfps> One problem with the axiomatic semantics is that it appears to me that it has a number of problems.
<trackbot> ISSUE-14 -- S14 might be about change not constraints -- pending review
pfps: I looked at it and it looks fine to me.
... I’m happy with the current status.
RESOLUTION: Close ISSUE-14
kcoyle: Simon is in the midst of adding the user stories
... There are now some new stories and some have changed
... We need to know which are approved
Arnaud: I think two more issues related to stories need to be resolved
<trackbot> ISSUE-15 -- S17 is about access not constraints -- open
Arnaud: I think there was no response from Dean on S17
... I will ping him
<trackbot> ISSUE-16 -- S18 does not appear to be about constraints -- open
pfps: David said it’s subsumed by previous story
... He said it may not be very constraint-like, but it seems like a useful facility
... Not a strong argument for enclosing it as a user story
<Arnaud> PROPOSED: ISSUE-16, dropping S18
RESOLUTION: ISSUE-16, dropping S18
[The resolution is to *close* ISSUE-16]
<pfps> As I've stated for some time now, I think that the document meets the (low) bar for FPWD, even without closing open issues or adding new stories.
Arnaud: Let’s move to Requirements
... I have a list of those that are under consideration. It needs updating.
<pfps> The revised version addresses my objection
cygri: This is the one I reworded and Jose made counterproposal
Jose: I proposed a simple rewording
Instead of “SHACL profile”, I would like “SHACL profile or core language”
<pfps> This would also be fine by me
<pfps> Unfortunately these fine points of wording seem to be causing problems - we have seen emails stating that the understanding of requirements is not shared amongst all WG members
<pfps> I think that Jose doesn't want his preferred solution to be *excluded* by the reading of this proposal
cygri: I think the wording “SHACL Profile” makes most sense and should address Jose’s concerns
Arnaud: We should have a separate discussion about the point of contention, but move on withe the requirement
ArthurRyman: There was a decision that there will be a high-level language that dosn’t require SPARQL, and a way to extend it
<pfps> This is getting into mechanisms for supporting the requirement, not the requirement itself.
... Ignore certain constraints if implementation can’t validate them?
<Arnaud> PROPOSED: change 2.11.7 per Richard's proposal amended by Jose
Arnaud: Let’s separate the discussion and talk about SPARQL later.
RESOLUTION: change 2.11.7 per Richard's proposal amended by Jose
<Arnaud> PROPOSED: Approve 2.11.7 Separation of structural from complex constraints
Peter’s objection is addressed, so we can approve this
<pfps> I'm not going to vote *for* it, but I'm not voting against it.
RESOLUTION: Approve 2.11.7 Separation of structural from complex constraints
Arnaud: We need to come to terms on this.
... One camp wants to start with the well-defined semantics of SPARQL, leverage this, build on top of this, and maybe define a simpler profile with higher-level building blocks
... You can claim conformance to the profile and don’t need to implement it in SPARQL
<Arnaud> An RDF vocabulary, such as Resource Shapes 2.0, for expressing these shapes in RDF triples, so they can be stored, queried, analyzed, and manipulated with normal RDF tools, with some extensibility mechanism for complex use cases.
Arnaud: And then there’s the other camp that wants to start with a high-level language
... The latter case seems to be more in line with what the charter says
... With SPARQL being an obvious candidate for the extensibility mechanism
... It may be that there’s not much of a difference between both options, where it’s just a question of which part one considers the core
... We should address this for real
pfps: Is your comment about the charter an official chair ruling?
... And if so, put it in writing so I can object to it?
... Those of us who think the SPARQL-based solution is superior would need to figure out if we need to push
Arnaud: To clarify, the charter does not exclude a SPARQL-based solution
Arnaud: I did not rule out the SPARQL-based proposal.
... As you know, WGs have some freedom in how they address the charter
ArthurRyman: Does the second camp say that we need to start from scratch?
... Totally new language and then figure out how it relates to SPARQL?
<pfps> It seems to me that the non-SPARQL camp wants the high-level language's semantics to be defined by something other than SPARQL.
Labra: I don’t think the approaches are incompatible. Let’s not say someone is in the SPARQL camp or not
... My point is that we should concentrate on the high-level language, and have implementations in different technologies
ArthurRyman: Rather than SPARQL just being an implementation language, I’d see it as defining the semantics of the high-level language
+1 to ArthurRyman
<pfps> +1 to Arthur
Labra: There could be a separate chapter that has mappings to SPARQL
... To define a high-level language, you need an abstract syntax
<pfps> There was something that Jose said that did not come through
Labra: I have a problem with a high-level language that’s only a single template construct
ArthurRyman: I agree with Holger that you don’t need an abstract syntax. The charter says we deliver an RDF vocabulary. That is the abstract syntax for me.
<pfps> The problem with just defining an RDF vocabulary is that there has to be a lot of stuff to say what is possible, and this is, in effect, an abstract syntax
ericP: At F2F we said there would be a language addressable through the RDF vocabulary and multiple semantics documents that say what it means, and we’d have a beauty contest
<Labra> +1 to pfps comment above
<ArthurRyman> agreed that an RDF vocabulary is not sufficient
<pfps> That said, the spec document doesn't need to talk about the abstract syntax per se
<ericP> cygri: jose said that the two options are to either have a high-level language or one that's built of blobs of SPARQL
<ericP> ... the actual proposal is that there be a high-level language defined using SPARQL templates
<Labra> I claim that just looking to the current draft of the spec
Labra: What I am saying is based on Holger’s spec
<pfps> An RDF vocabulary is a horrible way to define a high-level syntax, but it's not an impossibility to define a high-level language on top of an RDF vocabulary
<ArthurRyman> 1) we use RDF as the syntax, 2) augment that with a clear spec, 3) define the semantics in SPARQL (possibly as templates - avoiding circularity)
Labra: The spec has the priorities wrong. It starts with the template mechanism.
... Regarding abstract syntax. The term may not be popular, but an RDF vocabualry is not enough.
... Concurring with Peter.
Arnaud: Holger said there’s already a clear separation between SPARQL and rest
... One of the first examples uses a SPARQL ASK query
... So is there a way to do the same thing without SPARQL?
... If there is, it should be shown. If not, there is no clear separation.
... This is just from a casual reading.
<pfps> That is one of my issues with the spec document as it stands - there is no proposed alternative to SPARQL - if SPARQL is going to be the execution environment then don't gild this lily by saying that there are other possible execution languages
hknublau: No other proposed language supports global native constraints
... The language is flexible with regard to execution languages
<pfps> I think that it is a bad idea to permit other execution models
Arnaud: To be fair, there’s a disclaimer that says SPARQL is used here but others possible
... the spec looks like it endorses only SPARQL
<pfps> It's not the structure of the document that needs work, it's the philosophy of the document
<Labra> I proposed a lot of changes to the spec that were rejected by Holger
<Labra> +1 to pfps comment
<hsolbrig> My question is whether a compliant SHACL implementation must include a SPARQL interpreter and, if so, how complete?
cygri: Talking about the spec is too early. Design needs to be clarified. This is easier by talking about something concise like the pfps proposal.
Arnaud: ericP worked on primer
<pfps> What are we supposed to be looking at?
Arnaud: Please look at it and consider if it’s ready for FPWD
<hknublau> we need more meeting time
<Arnaud> pfps, the primer: http://w3c.github.io/data-shapes/data-shapes-primer/
<hknublau> another week went by with too little progress IMHO
<pfps> +1 to holger
<Arnaud> ok with me, we could add 30mn to the call
<hknublau> yes please
<Arnaud> I'll make the proposal to the list
I think we need some PROPOSALs on key questions, and focus on the conceptual level of what we’re building, rather than on what to call things and in what order to present things in the spec
<Arnaud> the only downside is that I may get cranky as I get more hungry :)
<Arnaud> or just crankier ;)
Cranky chairs are not always a bad thing!
<Arnaud> I think the way things are presented is very telling of what people's assumptions are though
<Arnaud> the discussion on core vs profile is a demonstration of that
Arnaud, that is true but you can make these assumptions clear much faster than writing a long spec
<Arnaud> we probably should have a few more highlevel issues raised to capture the pain points
<Arnaud> people can then make proposals to resovle these
Arnaud, is it productive at this stage to send PROPOSALs to the list, such as “PROPOSAL: SHACL must support embedded SPARQL in at least one profile”?
<Arnaud> you can, I think nobody will disagree with that
<Arnaud> the disagreement seems to be about what is considered the base
Well, but then the next is, “PROPOSAL: All high-level constructs must have normative definitions in SPARQL, except where that is impossible”
<Arnaud> do you take a top-down approach or bottom up
And: “PROPOSAL: SHACL includes a facility for defining new high-level constructs using SPARQL”
“(in at least one profile)”
<Arnaud> that's probably more controversial
<Arnaud> but feel free to ask and we'll see
Arnaud, right. But I think it would be helpful to establish some firm ground around the controversial questions
<Arnaud> I agree
<Arnaud> the truth is that we've tried to avoid some of these pain points to try and make progress (before people start jumping on each other's throat :)
<Arnaud> but obviously we can't keep avoiding them and need to resolve them at some point
<Arnaud> I think it has been a useful exercise in that it has allowed people to know each other and get to understand each other better
Yes. I think we know quite well where the fault lines are, and we perhaps need new input/opinions/voices for those that have so far kept out of these discussions
<Arnaud> which is key to having a productive discussion
<Arnaud> like who?
<Arnaud> oh sorry, I need to go
<Arnaud> you can type and I'll see it later
Ok will do
The SPARQL question has been dominated by the TopQuadrant folks and Peter and Jose. I’m not sure where Harold or Arthur or Karen or even Eric fall on some of these questions. If there are proposals with support from the majority, and strong opposition only from one or two voices, then at least it puts pressure on them to propose a way forward.
Anyway, dinnertime for me.
<Arnaud> trackbot, end meeting