See also: IRC log
scribenick pfps
<Arnaud> scribenick: pfps
arnaud: many people missing
kcoyle: I can do a report from a meeting last week
arthur: I talked with Holger and we made some progress
<Arnaud> PROPOSED: Approve minutes of the 19 November Telecon: http://www.w3.org/2015/11/19-shapes-minutes.html
minutes looked OK
RESOLUTION: Approve minutes of the 19 November Telecon: http://www.w3.org/2015/11/19-shapes-minutes.html
<aryman> thx to Arnaud for collecting the Resolutions
arnaud: the page on resolutions is now in reverse chronological order
arnaud: virtual F2F is two weeks from today
... timing will be midway between US Pacific and European time (Newfoundland)
... I want to tackle the hard issues
... If anyone has such please bring them forward
arthur: big issue is difference between SHACL and ShEx
pfps: how should we bring these up?
arnaud: wiki page section OK?
pfps: works for me, but we should point to that from the WG main page
... issues SHACL vs ShEX, recursion, UI, implementation vs specification
<kcoyle> makes sense to me, too
arnaud: some of this is meta-level discussion not particularly related to a single issue
... a decision on doing UI at all would have effects on several issues
... F2F schedule is currently on Newfoundland time (-3.5)
... there is no reason to cut off the last day for travel reasons because there will be no travel involved
... making three equal days means later start and earlier end, which makes things easier for everyone
kcoyle: +1
pfps: +1
<aryman> +1
arnaud: three six-hour days would give the same overall time
... I'll adjust the schedule accordingly
this means at 6am PST start and an 8pm CET end
kcoyle: SWIB meeting - Semantic Web in Bibliotech(?)
... I gave a short talk on SHACL that had quite a bit of interest
... There is someone working with SHACL from this community
arnaud: tracker had only a very few "products"/categories
... I added a few new ones to better characterize some of the issues
<aryman> 1+
arnaud: other WG members can go ahead and add new products and associate issues with products
... comments?
arthur: is there any connection between products and deliverables?
arnaud: it makes sense to have deliverables as products, but informally making sub-products helps to distinguish different bits of work
arthur: I like the splitting out of SHACL-SPARQL
... I'd like putting the SPARQL stuff into a separate document
... splitting has been discussed in the past, holger thought that that would diminish the importance of SPARQL
arnaud: TQ was worried that document splits might negatively affect the specification
... I would like to defer this and focus on the issues for now, though
... feel free to add new products, particularly SHACL- products, for major components of SHACL
<TallTed> there is no connection required/mandated between products and deliverables, though this is typical.
<TallTed> sub-deliverables being treated as distinct products has helped in other groups.
arnaud: there are a number of people who are not on the call and have not sent regrets
... Eric in particular said that he would talk about ShEX
... having the ShEX people more involved with SHACL is desirable
... Eric has said that there is an extensive test suite for ShEX that can be transformed into a test suite for SHACL
labra: we have been working on ShEX, I am not prepared for today because Erica was going to be talking
... we have a test suite, which can be used for some parts of SHACL
... I have been comparing the expressiveness of SHACL and ShEX
... I have a page on the comparison
... we have a paper on the comparison
<Labra> https://github.com/labra/ShExcala/wiki/ShExC-vs-SHACL
labra: ... is the page on the comparison between SHACL and ShEx
... one significant difference is a property that has a property that is optional but if present must be a particular value
... the paper also has a preliminary performance comparison
arnaud: great work
... it might be more productive to do this sort of work collaboratively
... it would be useful to get information out to the working group before the F2F so that WG members can prepare
<Labra> http://shex.io/
labra: link to primer on ShEx and other information
pfps: i don't find the preparation of a paper comparing ShEx and SHACL particularly heartwarming
arnaud: sharing between ShEx and SHACL sooner would be more helpful, I think
labra: this paper started several years ago, on ShEx only
... I was asked to compare with SHACL
... the performance results were a surprise to me
... ShEx performance is quite good compared to SHACL at the moment
... I have been sending information into the working group mailing list
arnaud: this is useful material for the working group so that it can make better choices for SHACL
... it does have the danger of being a pissing contest between ShEx and SHACL
... let's make sure that the working group can benefit from this
labra: this paper draft has just become readable
... the spirit of the paper is compare and propose
arnaud: there are three issues on the agenda that could be discussed, but there are few people on the call today
arthur: jose says that he has written a paper
<Labra> http://www.semantic-web-journal.net/system/files/swj1260.pdf
labra: the link to the paper is on the paper
... it has been submitted
arthur: can you put a pointer on the WG wiki
labra: it is under review, but I could do so
<aryman> Other Input Documents
arthur: the appropriate section would be other input documents
<aryman> on https://www.w3.org/2014/data-shapes/wiki/Main_Page
<Labra> Another paper: http://www.epsiplatform.eu/sites/default/files/Linked%20Data%20Validation%20and%20Quality.pdf
arnaud: we don't have enough people on the call to close technical issues
... is there anything to be discussed?
arthur: what is the proposal for issue 23
issue-23
<trackbot> issue-23 -- Shapes as classes -- open
<trackbot> http://www.w3.org/2014/data-shapes/track/issues/23
arnaud: this has been discussed at length
... there are two possible views on modelling using SHACL
... holger has backed down quite a bit
... his proposal is to not have shapeclass but to have an implicit looping scope link when a shape is also a class
arthur: this is just a shortcut, so I'm in favour
pfps: i find shapes as classes as morally repugnant
... I worry about objections from RDFS people
tallted: why are we worrying about future objections, that's just picking fights
... we are at a point were we need to make the best compromise we can come up with and let the objections come forward as they may
arthur: why does pfps find this morally repugnant?
... i think that we need to tighten the proposal up to determine how it works, e.g., when does this kick in
arnaud: this implies some inferencing - when does this stop?
pfps: SHACL does inferencing so we have already gone down that slope
... one of my problems with SHACL is that its inferencing is different from RDFS but on the same vocabulary
... my moral issue is that this is yet another push into modelling
arnaud: this is the core of the disagreement
pfps: it is likely that there will be objections no matter what happens here
arnaud: it is now the time to come up with a compromise and let the objects happen
arthur: we can keep out of trouble if we don't say inference, i.e., we don't talk about adding a new triple
... similarly transitive closure of subclass doesn't add new triples
pfps: that's abusing the work inference, you are still doing inference
arthur: there are no extra triples, so there is nothing that any other tool can get at
arnaud: we could have a section that precisely lays out what ¨inference¨ is being done
tallted: inference means adding a triple so if you never articulate the triple then you are not doing inferencing
arnaud: I bet holger wouldn't care how the wording is
pfps: in RDF and RDFS inference is not tied to adding new triples
arnaud: whether the triples are materialized or not is a separate question
arthur: when you run a reasoner you see extra triples being added to the dataset
pfps: that's a particular way of doing inference, there are lots of other kinds of inference
tallted: we don't use inference in the usual sense
... a shapes language is a modelling language so I don't understand the objection
<TallTed> tallted: what else is a modelling language, but a way to describe a bunch of shapes? what else is a shape description language, but a way to model a space?
arnaud: I hope that we can can come up with a way forward that is acceptable
... hopefully Eric will present next week
pfps: if Eric is going to present next week, can we some materials beforehand
Arnaud: I don't know that he has slides per se but I will ask Eric
<Arnaud> trackbot, end meeting