Meeting minutes
Drafts and publications
caribou: Publication process - start editors working drafts
… need WG decisions for these
… SHACL-Core and SHACL-SPARQL
Carine: Goal is to publish Core and SPARQL specs as First Public Working Draft within a month or 2
Editors
Eliana: For Core and SPARQL we have volunteers
Nick: Editors can go ahead and start editing
… we have enough people to start
<Zakim> betehess, you wanted to say we are interested and to say we Netflix are interested in the Profile spec
Betehess: volunteering for the profile spec
… depends on what the group wants to put in there
Youssouf: proposing to help editing Core and also profiling
<nicholascar6> I have added YoucTagh & betehess to the Deliverables they've indicated in the issue
Eliana: help is welcome on profiling
Carine: no limit in number of editors
<YoucTagh> Thank you
Nick: Editors can communicate directly between them
<betehess> Where are those scopes being discussed?
Nick: I'm happy to fill in
… I will forward to editors what we have discussed
AlexNelson: Will the issue describing the scope (of the profiles document) relate the scope to the SHACL-AF document?
<nicholascar6> I will create an Issue for the Profiles scope
Inferencing
Holger: the plan is to take care of Core and SPARQL to avoid delays
… but there's early interest in Inferencing, so it would be wise to start discussing
<nicholascar6> Scope of Profiles Spec Issue created: w3c/
<holgerk> https://
holgerk: rules making triples and derived properties - both are "inference"
… table count example - counts a path expressions
[Holger explains the "derived properties"
holgerk: another example - properties that cause a condition to be checked
… many customer examples
… computed on-demand e.g. GraphQL request
… value is always up-to-date
… I hope to see this in the spec in a more prominent way
… Computed targets. More ways to give target focus nodes.
… calculate a set of nodes can provide this
… (extreme case) dynamic constraints - experimental
… example - minCount is has a computed value at the time of constraint checking
… happens on the shapes graph from the data graph
… complicates the language - a tradeoff
… another example: "states code" depend on the country
… we already have some kind of inferencing in SHACL e.g. path expressions
… use blank nodes for path expressions to show they are calculated and not e.g. for a form.
… design pattern - blank nodes lead to calculations - an extension point in Core
… needed to give SHACL-SPARQL
… currently "node expressions" - other names possible
… a separate library document has the defined expressions
… generalize property paths to all any node expressions
… calculated properties can be ordered
… later work has inferred triple
… in addition to sh:path, extend sh:targetNode
<ajnelson-nist> Determinism question - could the order in which shapes node expressions are encountered (their shapes parsed) cause different computation results? Should this be allowed?
holgerk: to be discussed - more difficult to do static analysis
<ajnelson-nist> Recursion question - what happens if a dynamically-computed property references another dynamically-computed property? How much do we need to control for infinite re-computations?
holgerk: has a learning curve
… inference and recursion
<ajnelson-nist> (Pardon - recursion question noted in Discussion section)
holgerk: define a SHACL profile without node expressions
AndyS: Inference means a lot of things, this is mostly calculations, expecting more inferencing
… Can you summarize exactly which are the extensions points?
ajnelson-nist: Recursion - will comment on GH
… recursion needs thinking about
Github issue for discussions of node expressions : w3c/
caribou: concerned about the scope ... XSLT example
… first version needs care
holgerk: I agree
… one way to scope this is try to cover SPARQL which is already an extension point
… same expressiveness
… rdf exchange syntax for SPARQL
elianaP: This is "Spin"?
<simonstey> https://
holgerk: old RDF description of a SPARQL query to introspec on e.g. features used.
nicholascar: rules ordering
… derived properties can rely on one another
… can materialize, then execute
simonstey: If SPARL/RDF , do we need SHACL-SPARQL?
s,SPARL/RDF,SPARQL/RDF,
… we have sh:order to layer computation
… what if the sh:order affects with analysis of dependency?
… in favour of derived properties
… entire node expressions may be too much for this spec iteration
holgerk: SHACL-SPARQL defines the node expression. Maybe all needed but need string based syntax.
<nicholascar6> AndyS indicated Node Expression discussion should be carried out at w3c/
Meeting Times
caribou: Proposal: alternating: 09:00 UTC, and 21:00 UTC
Alternating every week, editors should make sure to add items to be discussed to the right agenda if they can attend only one of the alternating meetings.