Re: shapes and classes: different

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

On 02/05/2015 09:37 AM, Richard Cyganiak wrote:
> Peter,
> 
>> On 5 Feb 2015, at 12:03, Peter F. Patel-Schneider 
>> <pfpschneider@gmail.com> wrote:
>> 
>>>> What is a class definition in RDF or RDFS?
>>> 
>>> I’ll answer by giving an example. Here’s the definition of the class
>>>  qb:Dataset:
>>> 
>>> http://www.w3.org/TR/void/#dataset
>>> 
>>> A pitifully small part of that definition is also written down in the
>>> associated RDFS file here:
>>> 
>>> [[[ void:Dataset a rdfs:Class, owl:Class; rdfs:label "dataset"; 
>>> rdfs:comment "A set of RDF triples that are published, maintained or
>>>  aggregated by a single provider." . ]]]
>> 
>> So some triples in some RDFS file that have the class as a subject are 
>> in the definition.  But which triples?
> 
> For the purposes of this conversations, the ones that affect the 
> to-be-defined notion of “validity” mentioned below.

How about just saying that some predicates are used to determine how to do
validation, and leave out any notion of definition.

>>>> What does it mean to include a shape in a class definition?
>>> 
>>> Here’s what I mean when I use these words:
>>> 
>>> 1. The shape is published in the RDFS file that defines the class 
>>> (assuming shapes can be expressed in triples).
>> 
>> Again, this depends on being able to determine which RDFS file defines 
>> the class.
>> 
>>> 2. If an RDF graph G includes a statement that claims a resource X
>>> to be a member of the class, but the description of X in G doesn’t
>>> fit the shape, then G is “invalid”, for some to-be-defined notion of
>>>  “validity”. The important thing is that a validator can detect the 
>>> condition.
>> [I'm assuming that you mean "includes a triple X rdf:type C" or 
>> something like that.]
> 
> Yes, thanks.
> 
>> So to determine validity one has to find the RDFS file that defines the
>> class.  How is this done?
> 
> So, G is “invalid”, for some to-be-defined notion of “validity”. Maybe 
> this notion involves that G is “invalid with respect to D”, where D is 
> the graph in the RDFS file. And how exactly one finds that graph maybe 
> doesn’t have to be answered by this WG.
> 
> In other words, one validates against a particular graph containing 
> definitions.

Well, one could validate against a particular graph, and this only makes
sense if the graph has validation information. This sidesteps any need to
define what a definition is.
> 
> That being said, I’d say that a reasonable place to look for the graph
> is by dereferencing the URI of the class. I think this follows from the 
> general framework of web architecture. The graph you get from 
> dereferencing the URI is “authoritative”, and philosophically speaking, 
> we get the best interoperability if everyone uses that one. But in 
> practice people will often use graphs obtained in other ways, for all 
> sorts of reasons. For example, the graph may need to be “patched” a bit 
> to work with their particular toolset, or is missing labels that are 
> needed for UI purposes.

I agree that this is a reasonable thing to do in some cases. However, there
are cases where the "canonical" place does not have what is needed.


> Best, Richard

peter
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1

iQEcBAEBAgAGBQJU07wNAAoJECjN6+QThfjz9ocH/RpxM4+1bBx9CBUAOB0pLLWd
UCeoBCOCh1yiduDQ9e5bTuBHCM0vtO5Ml7fWdMr34wWM+9d/b/gR8m+draGocPaM
QeZr37u1PeoN3XS4wnR4Ykzls8wBt2UCt31Ih2E/NZAHJJix9YNNP0Fr/VYO5qdK
CzrfTRyeV+uCYGUik28hpyzBfCxs6Ta6Q6Byv2i6M0sVUDrN0dklr2GFeVJyVamr
rFJzYHKVOWPjCV0/yuqoYtbZFc4bDcO7whRDWfOdC/JNz8Wo90SXVoWgDS+f2SjO
C3rKqFVZuwssUctWTwSVEj5ph+2pjompmo4eN+MWV3fwEftI/lvh3tsvGf2CGN8=
=vvd8
-----END PGP SIGNATURE-----

Received on Thursday, 5 February 2015 19:06:06 UTC