Meeting minutes
<AZ> Welcome, Zakim
<Timothe> but I have trouble with zoom
<AZ> pchampin: you can edit by usign
<AZ> ...s/usign/using/
<gkellogg> chair?
ora: I suggest we start to make concrete plans what has to happen for the spec to go out one of these days. As co-chair I hope this WG can conclude its business rather rapidly.
<pchampin> our charter says 2 years, anyway :)
ora: not 3 years like RDF WG did. More in a year if possible. We have a lot of work ahead of us. There are big issues and many little details that need to be taken are of.
… for myself the biggest thing is defining the semantics of all this.
Souri: I could present a few slides that I prepared.
ora: let's do that first and then discuss our concrete plans going forward
Souri's presentation about "parallel edges"
Souri: Presents why RDFn should support explicit naming. He shows examples for representation & query in RDF-Star and RDFn
https://lists.w3.org/Archives/Public/public-rdf-star-wg/2022Dec/att-0004/RDFn_WG_Slides_v2.pdf
<AndyS> Previous related comment -- "grouping relation properties" : Tim Finin : June2020: https://lists.w3.org/Archives/Public/public-rdf-star-wg/2022Dec/att-0004/RDFn_WG_Slides_v2.pdf
Enrico: General comments: There is a major missunderstanding, we cannot have two parallel edges to the same label. It would be different instances. The fact that they have the same name is just a modeling problem.
… I don't fully understand what this reification is about. Is this about "these facts are true" or "I want to say that these facts represent an event in the real world and this event has properties.
… I don't have the history, that's why I ask.
Souri: Property graph is supporting it and we need interoperabiliity between RDF and PGs.
<Enrico> In property graphs edges are *instances* of the edge type, therefore have differnet labels
Souri: The popularity of PGs is a confirmation to why this is useful.
… From a practical POV it is important that we can present something that is simple enough to the consumer so it is accepted.
olaf: The one issue I have is that you show us user-facing syntax. This is not really RDF-Star. It has an abstract syntax
… I think this was a mistake in the CWG. We had discussions about user-facing issues. We should more often also focus on the abtract syntax.
… My question for the user-facing syntax would be what is the corresponding abstract syntax.
<Enrico> Sure, your can interperate with PGs with RDF-* is you enforce to have different labels for properties (e.g., knows1, knows2) even when they are subproperties of the same property (eg., knows): knows1 subpropety knows. knows2 subproperty knows.
<Enrico> is --> if
Souri: From a semantics POV (syntax might be slightly different) it is important to say that we can have multiple triples (or quads) that are identical S-P-O but they have their own existance. The implementation needs to support that semantics.
… they have distinct names and we need to be able to address them separately. That's the whole idea.
<pchampin> point of order: as much as I enjoy this discussion, we have 10 minutes left, and we need to settle on a recurring slot for the next meetings
Souri: we can talk about the abstract syntax, I'm not very familiar with it.
… I can have a look at it and propose what would need to change.
… If I say S-P-O it points to the same object all the time. That makes it hard to support what I propse.
… I really learned this from Jesus Barasa. He gave a presentation in 2017 where he said there are two problems between RDF and PGs. RDF cannot do edge properties and RDF cannot do multi-edges.
… I saw RDF-Star does the first but not the second. If we want interoperability we need that.
<Enrico> with reification you can have properties of edges. But edges shoild be uniquely identified.\
<Enrico> So RDF-star is enough!
Souri: The RDF-Star proposes adds an explicit thing but do we need that? From that POV I tried to come up with what I present with RDFn. I really want to subsume PGs, that noone can say you cannot do that in RDF.
<Enrico> (one issue could stem from opacity of reification)
Souri: Keep it as seamless as possible, that's my motivation. I will go more into the abstract part and see what it means.
ora: I'm not fully convinced that this is a question of naming. Currently S-P-O is unique. My worry here is that if we touch that, how far do the ramifications go if we say RDF-Star is not a set of triples. How much of the world actually breaks if we would do that.
… That's what I think we should really pay attention to.
doerthe: For me it feels that you mix syntax and semantics in some way. A triple will always be a triple, how can it have different meanings. Would every triple come with a name or how do you see that.
Souri: The uniqueness is with S-P-O-n. It's like twins, S-P-O is the same but they are not unique. there is n1 and n2.
… If you do not use explicit naming the uniquenes stays the same, S-P-O based.
… If you need parallel edges, you need the n. And this is the uniquenes.
… We have to have a way to represent that work in RDF. PGs do that today. S-P-O-null & S-P-O-null would be the same. n is needed if this is not enough.
Enrico: I think this is a non-issue. The proper name is to give unique names to the edges. I'm in the standartisation group for PGs as well. The way they do it there is exactly like this. It's the same label but there is a second instantiation in the system.
Ora: we need to understand how much leeway we have before RDF semantics breaks that we cannot accept. If we break RDF sematnics in some bad way we will never be done and it will be detrimential to RDF-Star. Let's continue the discussion per email.