The W3C JSON-LD Community Group

Go Back

W3C Logo

JSON-LD CG

Minutes for 2023-10-04

Present
Gregg Kellogg, Ted Thibodeau Jr., Anatoly Scherbakov, Niklas Lindström, David I. Lehn, Pierre-Antoine Champin, Benjamin Young
Chair(s)
Gregg Kellogg
Scribe(s)
Gregg Kellogg, Benjamin Young
Agenda
https://www.w3.org/events/meetings/1ab7df94-bb06-440e-a6b9-bc9022018c57/20231004T120000/
Topics
  1. Announcements and Introductions
  2. YAML-LD
  3. JSON-LD Issue Discussion
  4. JSON-LD python libraries (PyLd)
Gregg Kellogg is scribing.

Topic: Announcements and Introductions

Topic: YAML-LD

Anatoly Scherbakov: I couldn't join last time, but I've been working on various YAML-LD things.
... [screenshare]
... I've made about four pull requests in the last month, mostly related to the test suite.
... Basic and Extended tests have been merged, so we now have a single test manifest.
... We've discussed using the extractAllScript flag to work with YAML streams.
... We need some more tests, for example to check for keys that are not strings, which are not permitted in JSON-LD, but are in YAML (not -LD)
... I have my own implementation in Python (python-yaml-ld) to run tests.
... For example, the expand function deviates a bit from the JSON-LD spec, which is more inline with Python coding standards.
... It uses PyLd for doing the heavy lifting.
... Doesn't do compact, quite yet.
David I. Lehn: You said you're using PyLd; does it still work okay? It's pretty out of date
Anatoly Scherbakov: Presently, to satisfy the test suite, I haven't encountered any issues.
... I did submit an issue some time ago. It was referenced recently in an issue for rdflib, where people were discussing removing the existing support and use PyLd instead.
... They're concerned that PyLd hasn't been updated in a while.
... Do you plan to continue supporting PyLd
Pierre-Antoine Champin: Last commit on PyLD Aug 6, 2020
David I. Lehn: We'd love to, but we really don't have the time as it's not something we use.
... On my todo list was to do some updates to match the safe mode from jsonld.js.
... Definitely looking for people to help support the library.
Anatoly Scherbakov: I'd like to help, if I can.
... Not sure, where to start, though. Maybe we can circle back to this.
Niklas Lindström: I mentioned that I have my own implementation, but I don't want to detract from PyLd. Still maintained by looking carefully at the spec.
... I'd like to make sure we have a good string interoperable test suite.
... I "transpiled" Python version to Java.
Gregg Kellogg: Curious about the YAML 1.2 version declaration.
Anatoly Scherbakov: In the industry, YAML is moved in multiple contexts and I've seen multiple YAML files, but never seen the version directive.
Gregg Kellogg: Best to get rid of the declaration.
Niklas Lindström: +1 For that

Topic: JSON-LD Issue Discussion

Subtopic: Base Direction.

Gregg Kellogg: Could obsolete existing i18n datatype and rdf:CompoundLiteral.
Pierre-Antoine Champin: +1 For how you framed it: a future version could obsolete the other forms.
... As a maintanence group, we can't make a normative reference to RDF 1.2 until it's in CR.
... We can't publish an update JSON-LD containing this until RDF moves to CR.
David I. Lehn: What's the story with C14N output?
... Mostly for language tag canon
... I think both old forms will be obsoleted.
Anatoly Scherbakov: My understanding of JSON-LD is vague, but if the industry accepts JSON-LD-star, will it mean that that will incompass those other things?
... RDF-star is a draft, and there is a remote time when it's considered ready.
... How does RDF-star play with OWL.
... I'm using named graphs to herd my data, but I haven't gotten into a situation where information from different sources.
Niklas Lindström: I wanted to +1 whet gkellogg said about the RDF-star direction.
... To the questions of quoted graphs, pchampin and I have been exploring the use of named graphs instead.
... Regarding OWL, I use it, but it may be somewhat outdated by SHACL.
... I hope it might come back into fashion. One of the aspects of named graphs is that, in practice, it is a way to isolate data.
... There's an aspect of what quotation is, but these are things being discussed.
... I hope they will converge that doesn't complicate things too much. Including provenance.
... I think this corresponds to how named graphs are used in JSON-LD.
Pierre-Antoine Champin: Let's not frame OWL vs SHACL as being competitors, although it is spoken.
... OWL2 was published in 2012, based on RDF 1.0. I think it's considered quite mature and still working.
... It works with RDF 1.1 pretty well. I hope that RDF 1.2 will still be "OWL friendly".
... The RDF-star WG needs to be careful to not break it further.
... No signs of a new WG anytime soone.
Ted Thibodeau Jr.: I wouldn't say that OWL is out of fashion. Many people find it overly complex for use cases.
... Nothing wrong with that. There are simpler pieces of OWL that are in wide-use.
... owl:sameAs has been misused because of misunderstanding.
... Other concepts are easier to use and quite useful.
... SHACL and ShEx are indeed orthogonal, but some people use it as an alternative.
Niklas Lindström: For reference, "another way" of using OWL: https://github.com/niklasl/ldtvm/
... The RDF-star WG should consider that OWL needs to be considered.
Pierre-Antoine Champin: +1 To all TallTed said
Niklas Lindström: +1
Anatoly Scherbakov: One of the most important use cases is related to YAML-LD. Using OWL, I can reduce the amount of data I need to enter manually.
... Type inference is pretty useful. SHACL and ShEx are more validation languages.
Pierre-Antoine Champin: To add to TallTed's point that some parts of OWL are more used than others: https://www.researchgate.net/publication/301094549_RDFS-Plus
Ted Thibodeau Jr.: I don't think it's off topic. It's useful to have cross-focus discussions.
... Type inference is a perfectly good use of OWL. The open-world assumption means that there is no enforcement in RDF.
Anatoly Scherbakov: I think that's one of the most important traits of linked data.
... Not practical to run against all the linked data in the world.
... But useful for small pieces.
Ted Thibodeau Jr.: +1 Exactly
I agree with what everyone has said. And, there is no real competition.
... The library world is working towards using RDF.
Niklas Lindström: OWL works with a controlled dataset. But a lot of the data is wrong :)
... The strength of RDF is that data has semantics.
... I advise implementors to at least support sub-cassing. Without that, they'll loose much of the power.
Ted Thibodeau Jr.: When we talk about knowledge; all you have to do is break it. Everything comes down to Air, Fire, Water, and Spirit as an alternative the Hydrogen, Oxygen, etc.
... To be eternally true, and given the lack of temporality, RDF has tripped up many people.
... Named graphs are important.
Niklas Lindström: +1 To that too!
Anatoly Scherbakov: +1!

Subtopic: rdf:JSON datatype w3c/rdf-concepts#65.

https://github.com/w3c/rdf-concepts/issues/65 -> Issue 65 Value space of rdf:JSON datatype (by gkellogg) [spec:substantive]
Pierre-Antoine Champin: The value space is something ideal (platonic); implementations will have some representation of the value space.
... As long as there is no change in the consequence of how it is defined, I don't really care.
... My feeling is that going from JCS to INFRA would have no practical change.
Ted Thibodeau Jr.: Changing value space *probably* fine, but if it has no practical effect, what's the point of the change?
Gregg Kellogg: Looking for some feedback on using I-JSON
Niklas Lindström: I speculate, as my own use of JSON literals is pretty literal. I think it makes sense to limit the space to I-JSON to improve interoperability.
... Maybe there are some odd parsers, but that doesn't bode well for interoperability.
Pierre-Antoine Champin: +1, I would expect that non-I-JSON data would have a lot of interop issues anyway

Topic: JSON-LD python libraries (PyLd)

Anatoly Scherbakov: I want to talk about how we can support this work.
David I. Lehn: We haven't had time to work on the code-base; I see there are 53 issues and 19 pull requests, which may or may not be relevant.
... I haven't had the time to update testing.
... I did need to do a bunch of update to jsonld.js, so similar fixes are probably needed in PyLd as well.
... Help is needed.
... There are also some performance issues.
... I think it's being used widely, but there are alternatives now.
Benjamin Young: I think it's clear that PyLd isn't critical for DB now, so it's dropped down the list.
... However, it is critical for others. It would be helpful to curate the issues and PRs.
... We can watch for things from this group that would be easier to vet.
Benjamin Young is scribing.
Gregg Kellogg: As an open source maintainer, PRs are nicer than issues
... but even so...they take review time
... so it might be useful is someone emerges as a fellow core contributor
... as a practical means of moving things forward
... I did that myself during the 1.1 stuff--but I was under contract
... and I know a few things about JSON-LD
Benjamin Young: Exactly, it needs more help to pull an open-source project.
... It doesn't need to stay a corporate thing
... I'm actually at DB in a developer-advocacy role right now.
Niklas Lindström: I've been thinking about how to work together more on these things.
... I'm not sure how close PyLd it is to the algorithms in JSON-LD 1.1.
David I. Lehn: I don't know, off hand, but the JavaScript and Python code we started with may predate the spec.
Niklas Lindström: I smoked out some bugs recently, and there are similar behaviors between Gregg's code and the playground.
... There are similar spec bugs
David I. Lehn: The spec and the code was built at the same time, so they may be a bit out of sync.
Niklas Lindström: I think context caching wasn't too diffucult.
1+
Anatoly Scherbakov: Regarding issue triage; I'll add that to my TODO list.
... I'm concerned about my availability, I'll try to look into it further.
... Maybe we could schedule time to discuss on the next call.
... I'd appriciate others also looking at those issues.
... Regarding the organization, I do think that being in the DB organization creates a perception of support.
... Maybe moving it to another organization would be a good idea. Perhaps. it could co-reside with a YAML-LD implementation.
Anatoly Scherbakov: Regarding developer advocacy; it makes me sad that there are so few developers that know what linked data is.
... Most developers have know idea. General advocacy on linked data would be useful.
... What can be used to interest more developers?
Benjamin Young: These are all topics which are near to my heart.
... I think it's unlikely that any libraries will leave the DB organization.
... It's hard to spin up a new organization.
... The ones that last have a very active developer community.
... But, we are looking at several different approaches to managing multiple open source libraries.
... In the meantime, we have an open contributor policy.
... Having me in a full-time position should be a good signal that DB is serious about this.
Niklas Lindström: I think that governance is important, and I've been distressed about how other libraries have languished.
... You can't be both stable and a test-bed. Adding governance is important.
... For anyone who has time, I'd like people to consider my implementation.
... I can't guarantee stability.
... the TrLd license is 0BSD, sort of an unlicensed.
0BSD
Niklas Lindström: 0BSD
Niklas Lindström: One of the things it has is its own transpiler, which allows us to get Java and JavaScript out of it, which has been useful.
David I. Lehn: Should that go into the homepage implementation list?
... I hope it can provide some inspiration.