Spurious normative dependencies

[pchampin]

I noticed here (and James actually noticed that before me) that RDF 1.2 Semantics normatively depends on RDF 1.2 Schema and RDF 1.2 Turtle. This seems unnecessary (at least as normative dependencies).

[pfps]

Semantics depends on Schema as the definition of what vocabulary is in RDF Schema. Semantics depends on Turtle as the language that is used to write some of the RDF graphs it uses.

Does this make the dependence normative? It's unclear.

[pfps]

It does look as if these should be normative, so my suggestion is to close with no change.

[pchampin]

I appreciate that Semantics uses Turtle in places that are not strictly speaking "examples", and could then be considered normative. So the normative dependency on Turtle makes sense, I guess.

Between RDF Semantics and RDF Schema, I see how both directions can make sense, but I would be more comfortable to have RDF Schema depend on RDF Semantics than the other way around. For this I have two pragmatic reason:

• if the prose in RDF Schema contradicts (or seems to contradict) the formal description in RDF Semantics, I would prefer RDF Semantics to have precedence
• if we want to move RDF Semantics to REC, I would prefer if RDF Schema was not holding us back

[pfps]

OK, the normative link to Schema will be turned into non-normative, based on a vote in the WG meeting on 20 November 2025.

[w3cbot]

This was discussed during the #rdf-star meeting on 20 November 2025.

View the transcript

w3c/rdf-semantics#163

pfps: Line item 5 is about whether semantics should depend normatively on Schema.
… historically it's been Schema that has been normative.
… issue is whether that's a good idea. Should it be changed?
… that is a WG decision.

ktk: Opinions on that?

AndyS: it's marked "proposed closing".

pfps: that was before pchampin ... [?]

pfps: I marked for closing. pchampin said something afterwards.
… I'll take that tag off.

pchampin: The RDF Schema defines the vocab. RDF Semantics provides some formal interpretations of the terms. Those terms have consequences as inferences.
… two sides of same coin.

<pfps> I'm voting conservatie on this - keep Schema normative - a change to informative should result in lots of informative marks in Schema

pchampin: If somebody has a problem with inconsistency between prose and formal description. I would be more comfortable with formal description having precedence.
… other reason is we're trying to publish semantics before schema.
… having dependency other way around would slow process. Not a good reason by itself, but other reason is a good one. This strengthens.

ktk: shall we vote about it? pfps, WG decision would take the form of a vote?

pfps: yes.

pfps: I'll write something which would make the change.

<pfps> PROPOSED: Make the RDF Schema document informative wrt what vocabulary is of interest and what its meaning is.

AndyS: want to talk about Turtle, not Schema. After proposal.

pchampin: I have an issue with pfps' proposal. Seems to apply to Schema as a document. Not what I was suggesting.
… not suggestion to make Schema a non-normative document.
… suggesting to not normatively depend on it.

pfps: problem with that is we have 2 normative documents as to Schema. There are Semantics and Schema. Both define things.
… which is more authoratative?
… Could keep Schema normative and say if conflict, defer to Semantics.

pchampin: that's what I was suggesting.

AndyS: exact wording of proposal I read as making whole of Schema a Note.

pfps: would rip out all normative stuff in Schema.

AndyS: framing of proposal would make it a non-normative document.

pfps: what would be a smaller step to remove normative link from semantics to schema?

AndyS: I thought idea of saying semantics document is authoratative would be sufficient.

pfps: somebody else needs to craft that proposal.

pchampin: if anything, Schema should normatively refer to Semantics.

pfps: as long as there's conflict resolution statement added, that would work for me.

<tl> looks good

<pchampin> PROPOSAL: change RDF-Semantics to *non-normatively* refer to RDF-Schema; RDF-Schema should then normatively refer to RDF-Semantics (giving precedence to RDF-Semantics in case of conflict)

<pchampin> +1

<TallTed> +1

<niklasl> +1

<tl> +1

<olaf> +1

<doerthe> +1

<pfps> +0.5

<ktk> +1

<enrico> +1

<gtw> +1

<AndyS> +1

<lisp> +1

RESOLUTION: change RDF-Semantics to *non-normatively* refer to RDF-Schema; RDF-Schema should then normatively refer to RDF-Semantics (giving precedence to RDF-Semantics in case of conflict)

pfps: when do the notes/issues saying it was discussed show up? afterwards?

pchampin: yes.

pfps: not in issue yet, but expect to see it soon.

pchampin: I may need to patch minutes. Will do what is needed.

<Zakim> pfps, you wanted to say that there needs to be a conflict-resolution statement

AndyS: relationship between semantics and turtle. I don't mind which way it's decided.
… if turtle suddenly changed, it wouldn't really change semantics.

ktk: what do you propose? we should vote for that?

AndyS: whatever the editor wants.

pfps: I'm uncomfortable making link non-normative. In places where the semantics is defined. Includes turtle constructs.
… if you don't understand how turtle works, you can't understand definitions.
… little bits of turtle. not a lot of it, but curies are included, so can't just include n-triples.
… prefixednames.

pfps: I am uncomfortable making the link non-normative.
… it is pushing the limit. Don't want to have somebody come back to say we have to change this.

<Zakim> AndyS, you wanted to talk about the turtle dependency after the proposal

AndyS: propose cancelling next week's SPARQL TF meeting.

---