Re: Core or Lite?

The workshop report says that “SPARQL plays a prominent role in how people tackle the validation problem today” and that “constraints checking can be performed using SPARQL quite effectively”.

However, it also says that “SPARQL queries cannot easily be inspected and understood, either by human beings or by machines, to uncover the constraints that are to be respected”, and therefore “SPARQL does not constitute a complete solution”. From this arises the need for a declarative approach to constraints, and this cannot be delivered by SPARQL alone.

Hence the consensus opinion that you quote: A declarative high-level vocabulary for the most common cases, with an ability to break out into something like SPARQL for the inevitable more complex cases that can’t be handled by the declarative solution.

But it is clear from the workshop report that a declarative solution alone would not satisfy many of the workshop participants, and there’s nothing in it to suggest that complex constraints on the level of SPARQL are uncommon or not typical.

We need both; the declarative vocabulary and the expressive fallback. Proposals that only address one (like Resource Shapes, and most of the descriptions of ShEx, and Peter’s original CONSTRAINTS proposal) are insufficient to address the charter.

Best,
Richard



> On 24 Mar 2015, at 15:51, Arnaud Le Hors <lehors@us.ibm.com> wrote:
> 
> I think we're touching on the very point of division in the group: whether writing complex queries using SPARQL is the normal/common case or not. From my point of view the workshop clearly didn't support that point of view and this is why the charter is written the way it is. 
> 
> Participants agreed that we should have a declarative mechanism a la OSLC Resource Shapes and, in recognition of the fact that there is only so much one can do that way, an extension mechanism should also be available to address complex cases which can't be handled declaratively. Here is how the report reads (http://www.w3.org/2012/12/rdf-val/report): 
> 
> There was consensus on the need for 
> 1.        Declarative definition of the structure of a graph for validation and description. 
> 2.        Extensible to address specialized use cases. 
> 3.        A mechanism to associate descriptions with data. 
> 
> Note that this doesn't mean that the extension mechanism is any less normative than the declarative one but it makes a difference as to whether the extension mechanism is the center piece (or as Richard put it "the most basic construct") or not. 
> --
> Arnaud  Le Hors - Senior Technical Staff Member, Open Web Technologies - IBM Software Group
> 
> 
> Arthur Ryman <arthur.ryman@gmail.com> wrote on 03/24/2015 05:50:14 AM:
> 
> > From: Arthur Ryman <arthur.ryman@gmail.com> 
> > To: Holger Knublauch <holger@topquadrant.com> 
> > Cc: "public-data-shapes-wg@w3.org" <public-data-shapes-wg@w3.org> 
> > Date: 03/24/2015 05:51 AM 
> > Subject: Re: Core or Lite? 
> > 
> > Holger,
> > 
> > All aspects of SHACL that are described by the spec are normal in the
> > sense that they define compliant behavior. You seem to be implying
> > that only Part 1 is normal. I don't understand what is gained by this
> > use of the term "normal".
> > 
> > I'd also like to clarify that I think we only need one RDF namespace.
> > Part 1 defines some of the terms and Part 2 defines the rest of the
> > terms.
> > 
> > If we are going to use the term "normal", let's agree on the meaning.
> > One meaning is say that the largest user group is the "normal" one. If
> > that is the case then we have clear feedback that the majority of
> > users will want a high-level vocabulary for expressing common
> > constraints. A smaller, more advanced, set of users will write
> > constraints in SPARQL, JS, ShEx, etc.
> > 
> > -- Arthur
> > 
> > 
> > 
> > 
> > 
> > On Tue, Mar 24, 2015 at 1:50 AM, Holger Knublauch
> > <holger@topquadrant.com> wrote:
> > > On 3/24/2015 15:17, Arnaud Le Hors wrote:
> > >
> > > Holger,
> > > What would constitute the "extension mechanism" in your view then?
> > >
> > >
> > > The macro facility could be regarded as an extension mechanism for the core
> > > vocabulary. Another extension mechanism is the ability to use other
> > > languages such as shx:javaScript. But writing complex queries (e.g. using
> > > SPARQL) is not an extension mechanism. Also, using pre-defined macros from a
> > > 3rd party template library is also not an extension. So the headline of that
> > > Part 2 should reflect this differently. That was all I wanted to point out.
> > >
> > >
> > > I have to point out that Arthur's suggestion happens to be very much in line
> > > with what the charter calls for:
> > >
> > > An RDF vocabulary, such as Resource Shapes 2.0, for expressing these shapes
> > > in RDF triples, so they can be stored, queried, analyzed, and manipulated
> > > with normal RDF tools, with some extensibility mechanism for complex use
> > > cases.
> > >
> > > I don't think it helps to ignore that and try to force people into
> > > considering what was meant to be an "extensibility mechanism for complex use
> > > cases" the "completely normal use of SHACL".
> > >
> > >
> > > I stick to my statement that it is a completely normal use of SHACL to
> > > include SPARQL queries. It is also completely normal for OWL DL users to
> > > rely on features outside of OWL Lite. In the draft, SPARQL is part of the
> > > official spec. For a large class of users, what you call the "extensibility
> > > mechanism" will even be the main feature of SHACL. This includes people who
> > > currently use OWL and just want to use SPARQL for the bits that OWL cannot
> > > express. This is how TQ customers have operated for many years and is also
> > > the least disruptive path to adoption if we want SHACL to succeed with
> > > current semantic web people.
> > >
> > > What we have right now in the WG is that some people believe they don't
> > > really need SPARQL support, and that the core features are sufficient for
> > > most use cases. That's good for them, although not backed by much empirical
> > > evidence. At this stage we have no idea which features will be most widely
> > > used. Claiming that feature 1 is more important than feature 2  (and call
> > > feature 2 just an "extension") is premature and makes it more difficult for
> > > the supporters of feature 2 to get heard.
> > >
> > > The wording in the Charter was in retrospect unfortunate but it was
> > > difficult to clarify all these nuances in a single short sub-sentence. Back
> > > then I have been very clear that I will object to any attempts to
> > > marginalize the SPARQL support, and I will continue to do this. I hope the
> > > group respects the point of view of the SPARQL camp in the same way that we
> > > all respect the point of view of those who don't need really SPARQL support.
> > > My draft supports both view points.
> > >
> > > Regards,
> > > Holger
> > >
> >

Received on Tuesday, 24 March 2015 17:30:42 UTC