Unlinking Core from SPARQL

[bergos]

In the ticket about sh:pattern, the topic about making the core specification "SPARQL-free" showed up. I think we need to clarify that point.

My understanding was that SPARQL-based constraints or SPARQL-based Node Expression should all go to the SPARQL Extensions specification. I didn't expect that we would get rid of references in the definitions. We should keep them and continue using them to avoid redundancy and diverge.

[VladimirAlexiev]

• I am in favor of removing SPARQL from the Core spec, as a "reference implementation", or even in examples.
  The reason is that SPARQL does not facilitate efficient SHACL implementation since from the point of view of the implementor, it's not declarative.
  As confirmed by @HolgerKnublauch, Node Expressions are more declarative (though SPARQLy things like sh:limit, sh:offset really push the limit on "declarativeness").
• SHACL's Own Profiles #216 discusses that instead of Core vs AF, we need more granular SHACL profiles, eg
  Standard 1.0, with Node Expressions, with SPARQL, ...

[afs]

Somewhere, there should be the SPARQL for every core constraint (non-normative).
I don't mind which document that goes in.

[HolgerKnublauch]

I think we should avoid unnecessary work and duplicate content. When we currently reference SPARQL and doing so avoids redundancies then I would continue to reference SPARQL from Core. We also reference RDF, for example.

@VladimirAlexiev even when it's not declarative, the spec benefits from reusing terminology and definitions from SPARQL. We cannot replace our use of SPARQL with node expressions because that would bootstrap SHACL with itself.

My vote here is to keep things as they are and continue to reference SPARQL.