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

On 3/09/2020 18:04, Olaf Hartig wrote:
> On torsdag 3 september 2020 kl. 13:14:27 CEST Holger Knublauch wrote:
>> On 2/09/2020 23:17, Olaf Hartig wrote:
>>> On onsdag 2 september 2020 kl. 13:55:46 CEST Andy Seaborne wrote:
>>>> [...]
>>>> I need to work though the details in your message but could I ask what I
>>>> do wonder what the value of having the distinction in the formal model,
>>>> compared with, say, a data design pattern "PG mode is SA where each
>>>> referenced <<>> triple is also in the data graph".
>>> The problem is that the assumption of which mode to use would not be
>>> explicit in this case. For instance, if I put a Turtle* file online, your
>>> client does not know whether I meant this to be considered in PG mode or
>>> in SA mode. By using << .. >> exclusively for SA mode and {| ... |}
>>> exclusively for PG mode, we can be explicit. My proposed formalism
>>> carries over this distinction to the abstract data model.
>> I very much support this idea of using the syntax to define a dialect.
>> If I understand things correctly then PG mode is a subset of SA mode,
>> i.e. the space of possible PG graphs is smaller than that of SA graphs.
>> So I would assume that SA mode also still needs to allow << ... >>>
>> syntax for the non-asserted triples. However, it seems that PG mode
>> could be defined as the subset of SA that can be expressed through the
>> {| notation, basically capturing what property graphs allow. This is
>> probably another way of saying what Olaf has worked out with his more
>> mathematical/formal spec?
> Yes, you can certainly see it this way. The outlined formalism captures SA
> mode as a basis, and then adds PG mode on top as an option for the parts of
> your data for which you want to use it (which gives us a mixed-mode model in
> which we can explicitly use both modes within the same graph, just like in
> your proposed extension of Turtle*). Additionally, the formalism can be used
> a) to capture restrictions of the mixed-mode model to obtain a PG mode only
> version and b) to define mappings that represent everything in SA mode only.

Thinking this just a little bit further, there is a potential for better 
names for these dialects.

If RDF* in the most general sense is SA mode then the PG mode could be, 
for example, called RDF+ aka RDF plus. It would be a bit like OWL Full 
vs OWL DL, or SHACL-SPARQL vs SHACL Core. Some tools will elect to 
support PG mode/RDF+ only.

Holger

Received on Thursday, 3 September 2020 22:30:40 UTC