– DRAFT –
RDF-Star WG biweekly meeting

Approval of minutes from the last two meetings: 1 , 2

pfps: minutes look fine to me

<ora> PROPOSAL: Approve minutes of 2024-06-27 and 2024-07-11

+1

<tl> +1

<Dominik_T7> +1

<ora> +0 (I was not there)

<niklasl> +1

<AndyS> +1

<Souri> +1

<gtw> +1

<ktk> +1

<doerthe> +1

<AZ> +0 (was not present last week)

<enrico> +1

<Tpt> +1

<TallTed> +0 (haven't reviewed)

RESOLUTION: Approve minutes of 2024-06-27 and 2024-07-11

Proposed resultion by Semantics TF 3

the proposal is at https://github.com/w3c/rdf-star-wg/wiki/RDF-star-%22minimal-baseline%22

ora: I was not at the meeting, can anyone give a summary

nicklasl: the proposal adds triple triple terms and defines well-formed triple terms

nicklasl: the triple terms are fully transparent

nicklasl: this was discussed last Thursday and there were no objections

nicklasl: rdf:reifies with concept reifier

<niklasl> https://github.com/w3c/rdf-star-wg/wiki/RDF-star-and-LPGs

nicklasl: this is compatible with the use cases we've seen except for the one about recording deltas to RDF graphs

nicklasl: this also caters to provenance of unasserted triples and other advanced usages

nicklasl: this also permits many-to-many reifiers

enrico: there was some discussion on syntax and whether triple terms have to be restricted to object position

<doerthe> mmm, but I doubt that we simply should not have the injectivity

enrico: each unique triple term has a unique resource

enrico: doerte mentioned that this might be a problem in extensions

andys: there are two levels of syntax - general syntax and well-formed syntax

<doerthe> @enrico you have a funny way of talking me into the well-formed fragment: You introduce a problem I don’t think we have and then create a solution. ;)

<doerthe> I agree that the well-formed semantics solves the injectivity problem, I just doubt that we need to create the problem.

<Zakim> gtw, you wanted to ask about relying entirely on transparent triple terms

gtw: several months ago doerthe had some examples where entailment produced many-to-many results

<niklasl> I think Enrico's page https://github.com/w3c/rdf-star-wg/wiki/RDF-star-and-LPGs addresses this?

<enrico> To understand why the many-to-many is not prblematic for LPGs, see: https://github.com/w3c/rdf-star-wg/wiki/RDF-star-and-LPGs

gtw: there are some places where many-to-one is preferred

enrico: see the link above showing that the minimal baseline works for LPGs

doerthe: my point was that a previous proposal had problems

Kurt: my proposal is relevant to this issue

Kurt: LPG style semantics has local predicates

AndyS: part of the transparency issue is that IRIs are transparent in embedded triples

AndyS: there are conventions on use of IRIs on the web

gkellogg: the many-to-many issue looks similar to the situation with lists

<AndyS> Previous discussion about use of URIs -- https://www.w3.org/2024/07/11-rdf-star-minutes.html#t01

<AndyS> https://github.com/w3c/rdf-star-wg/wiki/Proposal:-Named-Node-Expressions

Kurt: [presents information from the slide above]

Kurt: a blank node is a local variable

Kurt: sometimes a local identifier is provided, but sometimes it is not, e.g., in [] syntax

Kurt: the blank node in a [] cannot be referenced elsewhere

ora: how does this relate to the minimum baseline

Kurt: if a local name could be added to a [] node then this would allow references to it elsewhere in the same document

Kurt: expanding the [] syntax to triples necessarily adds a local name for the blank node

ora: but if this name is an IRI there is a difference

<TallTed> Confusion may arise because `[]` are brackets, sometimes called square brackets. `{}` are braces, sometimes called curly brackets.

+1 to TallTed

I've tried not to use "brace" or "bracket" in scribing

<enrico> https://lists.w3.org/Archives/Public/public-rdf-star-wg/2024May/0067.html

enrico: I read the document and replied that this looks like singleton properties

ora: how would I know that this is like a singleton property?

I'm not geting the gist of this discussion so the scribing may not be accurate - see the document for more information

<Souri> I normally try to avoid bringing the complexity of blank nodes in discussions to avoid missing out on the elephant in the room. :-)

<enrico> This is where you can find the document by Kurt: https://lists.w3.org/Archives/Public/public-rdf-star-wg/2024May/0064.html

Kurt: in LPG there is an internal identifier for each edge, different from the label of the edge

<gkellogg> Sounds like for this to work as intended, the abstract syntax would need to allow blank node predicates.

ora: singleton properties have not gone anywhere in RDF

Kurt: my contention is that you do not have the concept of a singleton property then you cannot have the many-to-one case

enrico: there was a long discussion of singleton properties and it was turned down

<niklasl> +1 to enrico

<tl> we will have to have the discussion what the rdf:reifies relation actually means. we will then have to re-visit instantiation, e.g. singleton prperties

TallTed: as I'm reading this is appears to me that the author of the data needs to know that there are these two marries edges when the data is created and that is not viable

<niklasl> +t to TallTed

Kurt: without named node expressions you cannot create a singleton property

Kurt: Neo4J has an internal hidden identifier for the statement that acts as a singleton property

<niklasl> I do get that if someone wants exactly the design that LPGs have, that's singleton properties (as I wrote a while back). But that does not handle lots of what we're enabling with rdf:reifies and references to triples. If there is opposition to enabling that, please speak up now.

ora: let's go back to the queue

gkellogg: what this boils down to in the abstract syntax is to allow blank node predicates

<Zakim> TallTed, you wanted to note that masked identifiers are implementation details; they are not part of the abstract data model; they do not really exist. seeAlso Oracle's invisible ROWID.

<tl> @niklasl i'm opposed to that

TallTed: invisible ids are not part of RDF and thus not portable

Kurt: the idea is to promote these to IRIs

ora: back to the agenda

<TallTed> nonymizing anonymous nodes makes them no longer anonymous ... in which case they should just be nonymized

tl: I think that the minimal baseline is a good basis but leaves some things open

tl: what does the reifies thing mean, for example

tl: and unasserted assertions

tl: so I'm in favour of adopting the minimal baseline

ktk: adopting is just creating something to compare against

doerthe: so proposals from now on are made against the baseline

ora: what is the proposal to vote on?

<ktk> https://www.w3.org/2024/07/12-rdf-star-minutes.html

<gkellogg> w3c/rdf-concepts#96

<gb> Pull Request 96 Change the section on Quoted Triples to Triple Terms. (by gkellogg) [spec:substantive]

ora: the Semantics group only resolved to bring forward the minimal baseline

gkellog: i created a draft PR for the purpose but all we need is a resolution to adopt the minimal baseline

<ktk> https://github.com/w3c/rdf-star-wg/wiki/RDF-star-%22minimal-baseline%22

<ora> PROPOSAL: Adopt minimal baseline https://github.com/w3c/rdf-star-wg/wiki/RDF-star-%22minimal-baseline%22

<gkellogg> +1

+1

<niklasl> +1

<ktk> +1

<enrico> +1

<doerthe> +1

<ora> +1

<gtw> -0 ; I remain doubtful that this addresses LPG use-cases

<TallTed> +1

<AZ> +0

<AndyS> +1

<eBremer> +1

<tl> +1

<Souri> +1

<Dominik_T> +1

<niklasl> (Noting that I favor well-formed, or at least restriction on triple-terms as objects only.)

<Souri> +1 to niklas regarding well-formed

RESOLUTION: Adopt minimal baseline https://github.com/w3c/rdf-star-wg/wiki/RDF-star-%22minimal-baseline%22

enrico: doubters should start an email discussion

<AndyS> The baseline abstract syntax has triple terms only in the object position.

ktk: discuss use cases next week

got to go now

<niklasl> ... or perhaps categorize some of what we have as "anti"-use-cases.