Re: using classes to control constraints

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

On 02/11/2015 09:19 AM, Irene Polikoff wrote:
> I am still failing to see a sufficient distinction. Especially since, as
> you say, languages where  a definition can be both a class and a shape
> are possible and exist. I would also add desirable to possible.

Do you want the result of the working group to be something like OWL?  I'm
happy to go there, but I'm not sure that others in the working group want
that level of complexity.

> <What would it mean to assert that an object belongs to a ShExC shape via
> an rdf:type link?>
> 
> Why is this question relevant? We are not standardizing ShExC but
> creating a new standard.

I was just using ShExC as an example.   If you want, I'll ask it without
reference to ShExC.

What would it mean to assert that an object belongs to a shape via an
rdf:type link?

> If you want to point out that there are some languages where classes and 
> shapes must be disjoint, I am sure it is possible to define such language
> and ShExC may be an example of one. [I don't know if it is].

I'm not sure that a must modality is relevant here.  Some languages (e.g.,
some database systems) have things like classes and things like shapes and
the two are disjoint.  Other languages (e.g., OWL) have things that are like
both classes and shapes.  Which way to go is up to the working group, but I
think that the working group needs to know the differences between classes
and shapes before making a decision.

> Irene

peter


> 
> On Wed, Feb 11, 2015 at 11:40 AM, Peter F. Patel-Schneider 
> <pfpschneider@gmail.com <mailto:pfpschneider@gmail.com>> wrote:
> 
> In RDF (not RDFS) class membership can only be directly asserted, by
> means of rdf:type triples.  In RDFS you can assert class membership and
> via subclassing infer membership in other superclasses.  However, this
> all directly depends on asserting class membership via rdf:type triples.
> 
> You never (at least as far as I can tell) assert membership in a shape. 
> Instead a shape provides recognition conditions that precisely determine 
> which nodes in a graph belong to the shape.  A shape implementation
> turns this determination into a procedural process.  You can then ask
> whether a node in a graph belongs to a shape.
> 
> So, although shapes and RDF(S) classes are indeed related to sets, and
> one can talk about the members of a shape and the members of an RDF(S)
> class, there are distinct differences between the two.
> 
> 
> In some modelling languages there are constructs that are both classes
> and shapes.  For example, OWL descriptions work this way.  You can assert
> that an object belongs to a description and you can also recognize
> whether an object belongs to a description.  For many descriptions, this
> recognition process does not depend at all on any explicit typing
> assertions.
> 
> 
> What would it mean to assert that an object belongs to a ShExC shape via
> an rdf:type link?
> 
> 
> peter
> 
> 
> On 02/11/2015 08:09 AM, Irene Polikoff wrote:
>> Peter,
> 
>> Could you clarify this more? I don't see the contrast you are trying
>> to communicate and I am not sure what you mean by "recognized".
> 
>> Class membership is asserted. However, it can also be inferred which I 
>> am interpreting the same as "recognized".
> 
>> I don't think membership in the shape is recognized if by "recognized" 
>> you mean that there is some computational process that decides whether 
>> something is a member of a shape or not. It is asserted in a sense that
>> a statement is made about the membership.
> 
>> In the end, however, they both have members.
> 
>> Irene
> 
>> On Wed, Feb 11, 2015 at 9:33 AM, Peter F. Patel-Schneider 
>> <pfpschneider@gmail.com <mailto:pfpschneider@gmail.com>
> <mailto:pfpschneider@gmail.com <mailto:pfpschneider@gmail.com>>> wrote:
> 
>> Classes are things that you assert membership in.  Shapes are things 
>> that you recognize membership in.
> 
>> I'm certainly open to other syntaxes.
> 
>> peter
> 
> 
> 
>> On 02/09/2015 09:19 PM, Holger Knublauch wrote:
>>> Hi Peter,
> 
>>> your email below seems to clarify how OWL Closed World would work.
>>> But I don't see a response to my questions at the end of my previous
>>> email in this thread (at the bottom here), especially on whether you
>>> would accept any other syntax than OWL at all.
> 
>>> Thanks, Holger
> 
> 
> 
> 
>>> On 2/8/2015 8:26, Peter F. Patel-Schneider wrote: In OWL constraints 
>>> for RDF, OWL axioms are used as constraints.  However, this doesn't 
>>> make RDF(S) classes be constraints.
> 
>>> You still create RDFS ontologies in the normal way, and constraints 
>>> don't have a role to play there.  Or maybe you don't have an
>>> ontology at all.
> 
>>> It is only when you want to validate some data that the constraints 
>>> play a role at all, and the constraints don't play the role of
>>> classes or even part of the description of a class.  You can have
>>> multiple constraint sets that employ classes from a particular
>>> ontology depending on just how your data needs to be.
> 
>>> Note in particular that if you need named shapes (a.k.a. closed
>>> world recognition) that these named shapes are only used for
>>> recognition, i.e., there are no type links that make individuals
>>> belong to these shapes.
> 
> 
>>> peter
> 
> 
> 
> 
>>> On 02/07/2015 01:35 PM, Holger Knublauch wrote:
>>>>>> On 2/8/15 12:44 AM, Peter F. Patel-Schneider wrote:
>>>>>>> I am very strongly in favour of having shapes be different 
>>>>>>> from RDFS classes
>>>>>> Hi Peter, would you mind explaining your statement above? Your 
>>>>>> original proposal to the WG was OWL Closed World, which 
>>>>>> re-interprets restrictions with closed world meaning:
>>>>>> 
>>>>>> ex:Class a owl:Class ; rdfs:subClassOf [ a owl:Restriction ; 
>>>>>> owl:onProperty ex:property ; owl:minCardinality 1 ; ] .
>>>>>> 
>>>>>> The equivalent in LDOM is:
>>>>>> 
>>>>>> ex:Class a owl:Class ; ldom:property [ a
>>>>>> ldom:PropertyConstraint ; ldom:predicate ex:property ;
>>>>>> ldom:minCount 1 ; ] .
>>>>>> 
>>>>>> Where do these approaches differ? If you would not accept the 
>>>>>> second syntax, do you have any other syntax than OWL that you 
>>>>>> would accept?
>>>>>> 
>>>>>> Thanks Holger
>>>>>> 
>>>>>> 
> 
> 
> 
> 
> 
> 
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1

iQEcBAEBAgAGBQJU25+CAAoJECjN6+QThfjzPAsH/1IXBpWd+evKCi7zfU/q4ecP
5xwKKE1wETokmMfPillipmQpL2+ftvJ6GT/LS5j9rU7Rf8KSxzZxRoOTmqtRJXd3
KlDRxc7jovEMG3Fv1lO+x6+2oIoOrYbZUproUjAuBYRSiimD69Cayq+qUYVQR8LS
kvn5pZggO/WHyhl/kX6ckm49wRNn/qoBEecDcCtMgheSn6oT1hEwReUO3ZT7Lxc+
Dheave7zACogzcSRlBeIIQjbNjZrVFiFGCdqbt7chuo501t4vg8Fdso/UvVzW7mT
Xsp6Wdzi5duf7PWmdoLs3TH/CSMj/cHIzrlFTnIsJC9zZ0qHpcCR1ciKoM+KOQ4=
=Ahvj
-----END PGP SIGNATURE-----

Received on Wednesday, 11 February 2015 18:30:00 UTC