Re: Presentation about reifiers and graphs

Hi Enrico,

Thanks again!

On Thu, Sep 19, 2024 at 6:12 AM Franconi Enrico <franconi@inf.unibz.it> wrote:
>
> More comments:
>
> Slide 3
> (akin to literals) ==> dangerous parallel, I’d delete it.

Why is it dangerous? I know you've had reservations about literals
denoting resources; but as they are defined (as a tuple of a lexical
string and an IRI, and as denoting a value in the value space of the
datatype this IRI denotes), I do think there is a structural parallel
(as opposed to the primitive(?) IRIs and bnodes). But perhaps that is
too superficial a parallel.

To me, this goes back to the question about what they denote. (A
constituent of the graph or a propositional atom of the model? I tried
to touch on that only lightly...)

> Slide 4
> ...you need to identify them as concrete... ==> ...you need to associate them to concrete…

How about "To describe such things, you need to refer to triples from
concrete reifying resources."?

> Slide 7
> I guess that the first triple is syntactically wrong in (my) current understanding of Turtle.

It is valid. It's just an illustration of the basic form of a blank reifier.

> I don’t really understand why you spend so much time on named graphs. This may lead to endless discussions we are not prepared to handle. I repeat my arguments to avoid discussions:
>
>
> In the above “encoding” of a named graph, shared blank nodes labels across named graphs with a dataset MUST denote the same resource across the graphs in an RDF dataset.
> On the other hand, datasets have two possible formalisations of semantics: a dataset may be interpreted as the union or as the merge of the participating named graphs, characterising exactly the sharing or not sharing of bnodes.
> In Trig, a blank node label represents the same blank node throughout the TriG unique document, i.e., blank nodes sharing the same label in differently labeled graph statements are considered to be the same blank node.
> In general, when named graphs in a dataset have different origin, you cannot assume that the same blank node label in different named graphs represents the same blank node.

Well, part of the task as I took it was to present something about the
difference between reifiers and (named) graphs. This has been asked
this many times, and we did touch upon it during the last TPAC, IIRC.

I didn't want to go into the union vs. merge details too much in the
slides, as I thought focusing on refiers being within one graph was
more to the point. But I will make an attempt (if I can find enough
time), certainly considering your wording (and the rest of the thread
on this list) to reduce or otherwise improve the latter part of the
presentation.

> My conclusion: these things need to be discussed, and we haven’t done it yet, and therefore I’d restrain to say that we can represent named graphs.

I agree that we need to discuss it, and I don't want to present
something contentious as if it was decided.

But does the presentation give the impression that we can? That is not
my intent. I have seen the name component of the named graph tuples
being used for a wide variety of things though, some of them being
what reifiers actually now do work for, in them having semantics, all
being within the same interpretation.

(I should perhaps cut these slides up into two decks; presenting the
first part on Tuesday, and have the remainder as a seed for
discussion, on Thursday or the following regular meeting?)

Best regards,
Niklas


> —e.
>
> On 19 Sep 2024, at 06:30, Niklas Lindström <lindstream@gmail.com> wrote:
>
> Dear all,
>
> Thank you all for the valuable feedback! I've read all replies
> (including those off-list) and much appreciate the help. (My lack of
> replying is due to lack of time.) Alas, I've not yet been able to
> collate and address it all (I was away on a conference), but I've
> updated the presentation with some few adjustments and additions (it
> is still at [1], with diffs at [2]).
>
> The goal is to introduce reifiers, and also to compare them to graphs
> and named graphs, since questions about their differences are
> recurring. I want this to make sense practically, without straying
> from formal definitions. As expected, there are some disputed points;
> and I appreciate all help in reaching these goals. I'm sure the
> presentation can be made shorter, clearer, and more inclusive.
>
> (If I sense there is enough rough consensus on its message, I'd be
> happy to present this at TPAC on behalf of the group.)
>
> Best regards,
> Niklas
>
> [1]: <https://niklasl.github.io/rdf-docs/presentations/RDF-reifiers-1/>
> [2]: <https://github.com/niklasl/rdf-docs/compare/main...dev>
>
>
> On Sat, Sep 14, 2024 at 1:28 PM Niklas Lindström <lindstream@gmail.com> wrote:
>
>
> Dear all,
>
> I've drafted a presentation for the purpose of introducing reifiers
> and comparing them to graphs. It's currently at:
> https://niklasl.github.io/rdf-docs/presentations/RDF-reifiers-1/
>
> Ora: you're of course free to take what you like and discard the rest
> for your TPAC slides.
>
> I tried to keep it on par with the baseline, but of course there may
> be wording in there still to be agreed upon. If you see fit I can make
> a PR for further collaboration in getting a version of it to
> https://github.com/w3c/rdf-star-wg/tree/main/docs.
>
> Best regards,
> Niklas
>
>
>

Received on Thursday, 19 September 2024 21:57:45 UTC