Re: Annotation syntax [was: SPARQL* test suite]

Hi Olaf,

On 9/7/20 3:51 PM, Olaf Hartig wrote:
> Hi David,
> 
> On måndag 7 september 2020 kl. 11:07:20 CEST David Booth wrote:
>> On 9/7/20 3:42 AM, Olaf Hartig wrote:
>>> On lördag 5 september 2020 kl. 18:19:40 CEST Jos De Roo wrote:
>>>> Am not really close to this discussion, but still, I tried to implement
>>>> things in N3 and it
>>>> appears to me that:
>>>> {:a :b :c} :p :o. could be SA mode
>>>
>>> No. This is a statement about a graph, where this graph happens to consist
>>> of a single triple (:a, :b, :c). In contrast, RDF* (no matter which of
>>> the two modes) is about making statements about individual triples.
>>
>> But this is an unfortunate limitation of RDF*.  It is much more useful
>> and general to be able to annotate multiple statements at once, as can
>> be done in N3.   I think this is an important limitation to correct
>> prior to any standardization.
> 
> I am not sure what exactly you mean by annotating "multiple statements at
> once." I assume you are talking about annotations for a set of RDF triples
> (which, by definition, is an RDF graph). 

Yes, that's what I meant.

> If that's what you want, then you can
> use N3 I guess. RDF* is not meant to be used for that; instead, RDF* is meant
> to be used in cases in which we want to have annotations on the level of
> individual triples rather than on the level of a graph as a whole. Think of it
> from the perspective of Property Graphs where you have the notion of edge
> properties (key-value pairs associated with a particular edge).  > To me, such an
> edge property is something different than a property / key-value pair that I
> may be able to associate with the graph as a whole (even if the graph contains
> only one edge). Similarly, annotations on the level of individual triples are
> something different then annotations on the level of a set of triples (at
> least for me). For the latter we have things such as N3, and RDF* is for the
> former.

Okay, I see the distinction you're making.  Apparently I wasn't paying 
close enough attention.

But if RDF* is more like the edge properties of property graphs, then 
essentially it is a baby step toward n-ary relations, right?  After all, 
the edge properties of property graphs are just a special form of n-ary 
relations:
https://www.w3.org/TR/swbp-n-aryRelations/#useCase1

Standardized n-ary relations are still notably lacking in RDF:
https://github.com/w3c/EasierRDF/issues/20

Thanks,
David Booth

Received on Monday, 7 September 2020 20:22:35 UTC