The W3C JSON-LD Community Group

Go Back

W3C Logo

JSON-LD CG

Minutes for 2023-05-31

Present
Anatoly Scherbakov, Gregg Kellogg, Pierre-Antoine Champin, Ted Thibodeau Jr., David I. Lehn
Chair(s)
Gregg Kellogg
Scribe(s)
Gregg Kellogg, Pierre-Antoine Champin
Agenda
https://www.w3.org/events/meetings/1ab7df94-bb06-440e-a6b9-bc9022018c57/20230531T120000
Topics
  1. Announcements and Introductions
  2. YAML-LD
  3. JSON-LD Issue Discussion
  4. Next call

Topic: Announcements and Introductions

Gregg Kellogg is scribing.
Gregg Kellogg: We have a tentative slot on Tuesday at TPAC.
Anatoly Scherbakov: Do I need to stay for the whole meeting?
Pierre-Antoine Champin: Currently, JSON-LD is scheduled Monday afternoon at TPAC, but still tentative
... VC is on Thursday and Friday, RDF-star on Tuesday.
Pierre-Antoine Champin: Solid WG is currently on Monday afternoon, which overlaps with JSON-LD.

Topic: YAML-LD

Anatoly Scherbakov: I missed the last meeting due to confusion with the calendar.
... I've been working on YAML-LD, slowly making updates to the spec.
Json-ld/yaml-ld#99
https://github.com/json-ld/yaml-ld/issues/99 -> Pull Request 99 #97: Document Extended → Basic conversion (anatoly-scherbakov)
Anatoly Scherbakov: Most changes so far are non-normative and discuss the extended profile.
... I've also started working on json-ld/yaml-ld#100. The purpose is to unify the theme of our exampltes.
https://github.com/json-ld/yaml-ld/issues/100 -> Pull Request 100 #94 Unify examples (anatoly-scherbakov)
... The examples don't have a similar subject, so I've been working on a single theme for examples.
... I need to go through the spec again to look for bugs.
... First is to get these PRs merged. Then, there are some repeating examples.
... Then a review to polish the spec.
Gregg Kellogg: Once the group issues a final CG report, it can be taken up by the JSON-LD WG.

Topic: JSON-LD Issue Discussion

Gregg Kellogg: There was discussion on the last meeting about merging PRs to update the EDs for future publication.
Niklas Lindström: I suggested an issue we might talk about. w3c/json-ld-api#565
https://github.com/w3c/json-ld-api/issues/565 -> Issue 565 The algorithm for processing a reverse term returns too early (niklasl) ErratumRaised
Niklas Lindström: The spec text has an early return, which is shouldn't do.
... This is what the distiller and playground both do, so it is in practice what's happening.
... I've changed my implementation. It doesn't result in changes to existing tests.
Gregg Kellogg: We need tests that cover this area. If we added a test for the current algorithm, existing implementations would fail.
Niklas Lindström: Both positive and negative tests would be useful.
... It affects expansion, compaction, and RDF variations.
Pierre-Antoine Champin is scribing.

Subtopic: JSON-LD-star

Gregg Kellogg: I felt that it was premature to add annotation in JSON-LD-star
... but I added it in Turtle-star and SPARQL-star
... I think that an annotation syntax is important, even if its behavior is still not quite determined
... There has been confusion in the market place with the annotation syntax in SPARQL-star
Pierre-Antoine Champin: I'm curious about confusion with SPARQL-star. The semantics should be orthogonal as long as RDF/SPARQL behavior is really syntactic sugar.
... We should follow the principle of "least surprise".
... The annotation syntax in the CG report was fine by me, although there are some corner-cases where it gives surprise results.
Anatoly Scherbakov: `{$Id: joe, rdfs:label: Hacker}: {probability: 0.5}` # is this something YAML-LD-Star could look like?
Niklas Lindström: I appreciate the annotation syntax; if you have quoted triples, it's likely that you would also assert them
... I'm trying to go through library use cases, but hard to find enough time.
Anatoly Scherbakov: Or say...
... There are cases when using unasserted triples, but that's different than how quoted triples work.
... We have graphic descriptions of some text, which might be automatically classified and we want to describe that process.
... We might have a suggestion for a classification, which would not assert the triple.
... The JSON-LD algorithm discussed earlier is an example of this kind of markup.
... Another case, in w3c/json-ld-api#565, there's something I've been reluctant to suggest.
https://github.com/w3c/json-ld-api/issues/565 -> Issue 565 The algorithm for processing a reverse term returns too early (niklasl) spec:substantive, test:needs tests, ErratumRaised
... It might only be something for JSON-LD-star.
... I have a single annotation, but the visual grouping makes sense.
... It would be nice to have a keyword, such as `@quoted` for grouping unasserted triples.
... What happens that it is opaque, but it might not look that way.
I use named graphs for similar purposes, and consider them unasserted.
... The thing with `@quoted` is that it might be a better form than having one triple in an `@id`, which is a problem for many use cases.
:Linkoping :population 1234 {| :source :wikidata |}.
|≠
:Linkoping :population 01234 {| :source :wikidata |}.
Pierre-Antoine Champin: The Turtle-star example does not entail the following because the lexical form of the integer is not the same.
... The asserted triple is entailed, but the annotation does not propagate to the inferred triple.
... The triple looks transparent, but it's also opaque, which is confusing.
... We need to balance expectations and ergonomics.
Niklas Lindström: I think ergonomics trumps a lot in practice.
... I think RDF is as stable as it is because of the formalism. I still think named graphs should play a role.
... I do have a suggestion for Turtle-star that should come up later.
... I used "%" to quote predicate/object pairs.
... That unifies the annotation syntax with quoted triples.
... The `@quoted` idea works well for JSON-LD, but may not work elsewhere.
Anatoly Scherbakov: As I understand it, it's effectively able to make annotations on graph edges.
... I have a YAML segment to do something similar.
Pierre-Antoine Champin: That case is what RDF-star is about. That example is interesting, but a bit scary for how it opens to some feature specific changes.
Ted Thibodeau Jr.: Because Zoom chat is not logged...
Ted Thibodeau Jr.: Example 1
Ted Thibodeau Jr.: ```
Ted Thibodeau Jr.: {$Id: joe, rdfs:label: Hacker}:
Ted Thibodeau Jr.: Probability: 0.5
Ted Thibodeau Jr.: Prov:wasDerivedFrom: joehacker.org
Ted Thibodeau Jr.: ```
Ted Thibodeau Jr.: Example 2
Ted Thibodeau Jr.: ```
Ted Thibodeau Jr.: <X> a :Organization {| :statedIn <v2> |} ;
Ted Thibodeau Jr.: % A :Person {| :statedIn <v1> ; :retractedIn <v2> |} ;
Ted Thibodeau Jr.: % :Employs <y> {| :suggestedBy <q> {| :statedIn <v2> |} |} .
Ted Thibodeau Jr.: ```
Gregg Kellogg: Syntax matters, but it has to flow nicely into the abstract syntax
... syntactic layers can keep people away from the triples, which could be dangerous
Ted Thibodeau Jr.: Example 3 -- "Quoted P-O-pairs" version:
Ted Thibodeau Jr.: ```
Ted Thibodeau Jr.: $Id: joe
Ted Thibodeau Jr.: $Quoted:
Ted Thibodeau Jr.: Rdfs:label: Hacker
Ted Thibodeau Jr.: $Annotation:
Ted Thibodeau Jr.: Probability: 0.5
Ted Thibodeau Jr.: Prov:wasDerivedFrom: joehacker.org
Ted Thibodeau Jr.: ```
Ted Thibodeau Jr.: Example 4 -- (lost to closed zoom)

Topic: Next call

Anatoly Scherbakov: Two weeks is good, it is stimulating
Gregg Kellogg: There has been some push-back on a two week schedule. It helps focus things.