W3C

– DRAFT –
RDF and SPARQL Working Group

20 August 2026

Attendees

Present
AndyS, doerthe, fsasaki, gtw, ktk, lisp, niklasl, olaf, ora, pchampin, pfps, rubensworks, Souri, TallTed, tl
Regrets
enrico, j22, Tpt
Chair
ora
Scribe
tl

Meeting minutes

Horizontal Reviews Tracking & Followup

<ktk> https://github.com/orgs/w3c/projects/20/views/5?filterQuery=is%3Aopen+is%3Aissue+-label%3Aaction+-label%3A%22use+case%22+-label%3A%22ms%3Afuture-work%22+%22Horiz%22

<ktk> https://github.com/orgs/w3c/projects/20/views/5 simply add "Horizontal" as filter

<ktk> w3c/rdf-n-quads#93

<gb> Issue 93 Horizontal Reviews tracker for RDF N-Quads (by pchampin) [ms:CR] [Tracking]

<ktk> w3c/rdf-n-triples#79

<gb> Issue 79 Horizontal Reviews tracker for N-Triples (by pchampin) [ms:CR] [Tracking]

pchampin: we might ping them again

<ktk> w3c/a11y-request#138

<gb> CLOSED Issue 138 rdf12-n-triples 2025-11-25 > 2025-12-23 (by ktk) [LC] [REVIEW REQUESTED] [pending] [s:rdf12-n-triples]

ktk: we should go through them and hint that they are very related

<ktk> w3c/security-request#110

<gb> Issue 110 rdf12-n-triples 2025-11-25 > 2025-12-23 (by ktk) [REVIEW REQUESTED] [pending] [LC]

ktk: security is still open, maybe because we lack a threat model?

<ktk> w3ctag/design-reviews#1161

<gb> Issue 1161 WG New Spec: RDF 1.2 N-Triples (by pchampin) [Review type: horizontal review]

ktk: TAG is also still open
… should we ping the TAG again?

pchampin: will check with Yves and Sarven

ktk: no feedback from security group on n-triples since 10 months

pchampin: probably depending on a better worked out threat model doc (to be provided by us)
… it's editorial work, probably not hugely complicated, but needs time...

ktk: we have to do threat model and internationalization to make progress on these
… timeline for RDF/XML
… can we alerady have a timeline for SPARQL?

andys: didn't have much time b/o other wg's

ora: what are we still missing in broad strokes?

andys: exists, variables, a lot of errata

rubensworks: for SPARQL would it be an option to set a deadline, and what doesn't meet the deadine would go into future work?

andys: maybe triage ->maintenance

olaf: some formalization issues open, blank nodes being one of them

<lisp> tomorrow +1

andys: meeting tomorrow, and we can do some triage

Review of open PRs, available at 1

ora: moving on to PRs
… what about w3c/rdf-common#21

<gb> PR 21 Add hljs configuration for TriG and SPARQL and a load highlighter utility function (by niklasl)

niklasl: yes

ora: does anybody review it?
… otherwise i can just go ahead and [make this go away]

andys: what means "vendorize"?

niklasl: using some vendor-specific, non-standard stuff
… that stuff is MIT-licensed

ora: objections to merging this?
… not hearing any objections, so suggest you go ahead and merge it

niklasl: maybe will improve the example file

<olaf> w3c/sparql-query#410

<gb> PR 410 Algebra: add an explicit "distinct" flag to Aggregation() (by Tpt)

olaf: ther is a PR w3c/sparql-query#410
… which is only missing Greg's review

gtw: back from holiday, will take a look

ora: w3c/rdf-turtle#157

<gb> PR 157 Move the reifiedTriple syntactic sugar explanation into the note (by domel)

andys: comment by pchampin from 43 minutes ago

ora: we'll wait for pchampin
… james, where are we with w3c/rdf-n-quads#108

<gb> PR 108 Undefined errors (by lisp)

ora: maybe redundant with n-triples?

lisp: if thesame text is amenable top all teh others i could add it there as well

ora: we approved n-triples, why wouldn't we use the same for all of these?

lisp: understood that to be the intent

andys: so we need to align the two texts

ora: james, can you make them all the same?

lisp: can do, will do

andys: needs care b/o TAG

<AndyS> w3c/rdf-n-quads#108 (comment)

lisp: will replicate the changes to all files

andys: that comment (above) needs consideration first

ora: james, please connect with pchampin

lisp: will do

ora: w3c/rdf-primer#24 darkmode
… niklas what's your take on it?

niklasl: think it can be merged

ora: a few test related PRs - any of them ready to merge?

andys: gtw, can you take a look at w3c/rdf-tests#356

<gb> PR 356 SPARQL: Set mf:specVersion on manifests and use IRIs (by Tpt)

gtw: will do

ora: what about w3c/rdf-tests#324

<gb> PR 324 manifest with rdf action model, revised with additional tests as per … (by lisp)

ora: greg, are you happy with this?

gtw: will have to have a look, seems big

lisp: is a side effect of a discussion in the SPARQL
… first version was generated by an LLM, later did let an LLM generate a requirements document and took that as input
… but the group should think about if/how LLMs can be used in the context of our work
… i made good experience with letting an LLM create requirements documents from specs
… that PR is just a result of such an approach

ora: are you saying that the discussion so far are only in the abstract?

lisp: yes

rubenworks: made soem changes

<rubensworks> w3c/rdf-tests#366

<gb> PR 366 Add collection point for RDF 1.2 implementation reports (by rubensworks)

andys: not comfortable with getting ahead to where W3C may go wrt LLMs

<TallTed> see Use of Large Language Models in Standards Work: W3C Group Note 24 March 2026

andys: the results are so depending on the prompt, they have too many inconsistencies, basing tests on them would add considerable risk

ora: can you give input to W3C

andys: maybe at TPAC

tallted: there is a related note (link above), just a couple of pages
… use of LLMs is neither blessed nor cursed

ktk: i do use it but i do see andy's point, but i also see james' point that there's only talk about talking about it. james, would you like to provide some input to the commitee?

lisp: would be easy for me to walk you through how 1.2 implementation was done, because it demonstrates a method which addresses one of andy's concerns and provides control

ktk: maybe we can have session where you demonstrate your approach?

ora: next week, this wg?

ktk: will put it on the agenda for next week

lisp: will be here

ora: maybe i'm a bit naive but i don't find this to be real rocket science
… except if your so naive to take the LLM output as fact

andys: exactly

ora: but you put _your_ name on it, not the LLM's output

tallted: very important that area experts control an LLM's output

lisp: key takeaway is that it's a search process and the quality of the result depends on the quality of the constraints the LLM is given

<TallTed> OpenLinkSoftware/ai-agent-skills

Minutes manually created (not a transcript), formatted by scribe.perl version 248 (Mon Oct 27 20:04:16 2025 UTC).

Diagnostics

Succeeded: s/Eve and/Yves and/

Succeeded: s/a lot of PRs/a lot of errata/

Succeeded: s/ [???]/maintenance

Succeeded: s/tool request/PR

Succeeded: s/Greggs/Greg's/

Succeeded: s/teh /the

Warning: ‘s/goddamn /https://github.com/w3c/rdf-primer/pull/24’ interpreted as replacing ‘goddamn ’ by ‘https://github.com/w3c/rdf-primer/pull/24’

Succeeded: s/goddamn /https://github.com/w3c/rdf-primer/pull/24

Succeeded: s|pull/24dark |pull/24 dark

Succeeded: s/vague/big/

Succeeded 5 times: s/james: /lisp: /g

Succeeded: s/rubenworks/rubensworks/

Succeeded: s/consdierable/considerable/

Succeeded: s/implementatuion/implementation/

Succeeded: s/demosntartes/demonstrates/

Maybe present: rubenworks

All speakers: andys, gtw, ktk, lisp, niklasl, olaf, ora, pchampin, rubensworks, rubenworks, tallted

Active on IRC: AndyS, doerthe, fsasaki, gtw, ktk, lisp, niklasl, olaf, ora, pchampin, pfps, rubensworks, Souri, TallTed, tl, Tpt