The W3C JSON-LD Community Group

Go Back

W3C Logo


Minutes for 2023-06-14

Topic: Announcements and Introductions

Gregg Kellogg: TPAC page up:
saumer: Working with cultural arts and Ruby in JSON-LD. [scribe assist by Gregg Kellogg]

Topic: YAML-LD

Anatoly Scherbakov: I haven't been able to do much of use recently. There is a draft PR with examples, but it requires more review. [scribe assist by Gregg Kellogg]
Gregg Kellogg: ... I think that the most uncertain thing in the spec is the conversion from YAML to JSON.
Gregg Kellogg: ... It seems that it is not standardized, particularly as there are multiple versions of YAML.
Gregg Kellogg: ... Do we need to formalize the process, as it is rather vague?
Pierre-Antoine Champin is scribing.
Gregg Kellogg: I don't think it is our role to standardize that
... there is a JSON profile for YAML, we could use that
... there are multiple ways from JSON to YAML
Gregory Saumier-Finch: Hi all.
... we, OTOH, are not going strictly from YAML to JSON, but from YAML to the internal representation
... there are not multiple ways you can do that
Anatoly Scherbakov: For the features we're using, we can consider it straightforward. [scribe assist by Gregg Kellogg]
Gregg Kellogg: ... YAML has a deprecated feature, where some integers are interpreted as base 60 rather than base 10. Version 1.2 cancels that, but it is still persistent.
Gregg Kellogg: My feeling is that we should stick to the JSON profile
Niklas Lindström: Is the YAML JSON profile official? (Is there a link?)
... (aside about CBOR)
... we may consider other profiles in the future
... but we should first look to solidify the base profile first
Pierre-Antoine Champin: About the question of profiles, there are two ways to look at YAML-LD: [scribe assist by Gregg Kellogg]
Gregg Kellogg: ... First would be a way of writing JSON-LD using YAML.
Gregg Kellogg: ... The other would be to make existing YAML data interpretable as JSON-LD. We can't then assume that that data will directly correlate with our data model.
Gregg Kellogg: ... This may help us figure out what to focus on.
Gregg Kellogg: I agree with pchampin that we need to spell this out. The second option may be too difficult in practice. There is still much JSON that can't be interpreted as JSON-LD, and JSON is simple.
Pierre-Antoine Champin: Just like JSON-LD doesn't cover 100% of JSON in the wild, neither will YAML-LD cover 100% of YAML in the wild. [scribe assist by Gregg Kellogg]
Niklas Lindström: Blind spots come from the data model being used; if it wasn't designed as triples, it can have all kinds of trouble. [scribe assist by Gregg Kellogg]
Pierre-Antoine Champin is scribing.
Gregg Kellogg: In JSON 1.0, we tried to hard to accomodate many JSON "dialects"
... let's focus on simple things first
... W3C now supports the "living standard" model
... we should aim for a simple First Public Working Draft, on which people can then comment

Topic: JSON-LD Issue Discussion

Subtopic: w3c/json-ld-api#566 -> Pull Request 566 Fix context processing for reverse terms (niklasl)
Niklas Lindström: Last time we talked about why to make the change. It solves w3c/json-ld-api#565, but I think remaining issues have been solved. [scribe assist by Gregg Kellogg] -> Issue 565 The algorithm for processing a reverse term returns too early (niklasl) spec:substantive, test:needs tests, ErratumRaised
Gregg Kellogg: ... I couldn't find any documentation about adding tests. I found something to add, but wasn't sure how to do this.
Gregg Kellogg: ... In this case, I figured that there might be many tests are round-tripable, and maybe we should make that explicit.
Gregg Kellogg: We usually duplicate expansion tests in the toRdf and fromRdf test suites
... because not all implementations of toRdf and fromRdf rely on an explicit expansion step
... this is burdensome...
... remember that it is a 10-years old test suite, with a lot of legacy
... adding instructions about how to add tests in the README would be a welcome PR :)
Niklas Lindström: The from- and to-rdf tests would be useful as well. [scribe assist by Gregg Kellogg]
David I. Lehn: The there are some overlap between expansion and toRdf tests, so you can't simply use the same test numbering. [scribe assist by Gregg Kellogg]
Gregg Kellogg: ... We had a mess in the past, and we've spent some time cleaning them up.
Gregg Kellogg: ... But, there are probably some missing tests.
Niklas Lindström: I didn't really see any obvious problems. If I devise an ambitious new naming scheme, we'll see ... [scribe assist by Gregg Kellogg]
Gregg Kellogg: ... In this case, as there's a standard, I'm not too troubled with duplication.
Gregg Kellogg: A matrix structure could be more appropriate than the manifest structure
... [discussion about the naming conventions for the tests]
... ideally testing infrastructure do not rely on test names
... once we achieve consensus on this PR, the question is "what next?"
... I think that we agreed that, given sufficient participation by the members of the WG, we could merge them
... leaving time to people to react

Topic: Open Discussion

David I. Lehn: The Verifiable Credentials group has a PR for doing data integrity, which includes JSON-LD contexts. [scribe assist by Gregg Kellogg]
Gregg Kellogg: ... It seems like something that the JSON-LD group may want to weigh in on.
Gregg Kellogg: There are groups out there doing cross-sectional work (e.g. i18n)
... JSON-LD is not one of them, and it would be strange for us to be the only arbitres for how to use it
... maybe we may add something to the Best Practices
David I. Lehn: Do we have a Data Integrity BP?
Gregg Kellogg: Data Integrity is not really related to JSON-LD is it?
David I. Lehn: People would argue that it is
... we had discussions in the past about using hashlinks, etc...
Gregg Kellogg: I think we found that hashlinks were the solution
David I. Lehn: Yeah... some people disagree with that
Gregg Kellogg: I think some groups should be chartered to deal with this.
... Not a job for a JSON-LD maintenance WG.
David I. Lehn: We could add small things that would make these use-cases easier.
... some properties that are metadata, part of the processing, rather than data
Gregg Kellogg: We looked at SRI (Sub-Resource Integrity) but did go far
... addressing it would need specific people
Gregory Saumier-Finch: Is the JSON-LD-star work exclusively part of this group? Or is there a relationship with the RDF-star group? [scribe assist by Gregg Kellogg]
Gregg Kellogg: JSON-LD-star is a product of the JSON-LD Community Group, strongly related to the work of the RDF-star WG
... it made more sense to keep it in the JSON-LD group than adopted by the RDF-star WG
... a lot still needs to be done, pending the progress of the RDF-star WG
... the annotation syntax is what is holding this work back, but the RDF-star WG is making progress on that front
Niklas Lindström: +1
Niklas Lindström: My recent thoughts:
Gregg Kellogg: Still time to look at alternative representations of quoted triples
Niklas Lindström: I pasted the link to my thoughts on JSON-LD-star. Maybe we can take it up in the next meeting. [scribe assist by Gregg Kellogg]
Gregg Kellogg: ... One of the things which is tricky is that people tend to want to use annotations to say things that are about the object, and not the assertion.
Gregg Kellogg: ... You could imagine a shortcut using OWL to talk about the object in the context of the assertion, which might be useful.
Gregg Kellogg: ... [More discussion of RDF-star semantics].
Gregg Kellogg: ... But, JSON-LD is about making things possible, and adding a checksum to a link makes a statement, which may not be captured semanticallly.
Gregg Kellogg: ... It's unclear about if I can go from my assertions to the paper where they were originally asserted.
Gregg Kellogg: ... [Allusion to Escher's picture of a dragon biting it's own tail].
Niklas Lindström: In Gödel, Escher, Bach, Douglas Hofstadter interprets the dragon's tail-bite as an image of self-reference, and his inability to become truly three-dimensional as a visual metaphor for a lack of transcendence, the inability to "jump out of the system"

Topic: Next call

Gregg Kellogg: Next meeting June 28th.
Gregory Saumier-Finch: Thanks
Niklas Lindström: +1 That's the one