Re: ISSUE-23: A specific proposal

https://www.w3.org/2014/data-shapes/wiki/Requirements#Evolutionary_Adoption
_Path

As for the other requirements, using rdf:type to directly link instances
with shapes is one way to implement them.




On 4/28/15, 8:28 PM, "Peter F. Patel-Schneider" <pfpschneider@gmail.com>
wrote:

>-----BEGIN PGP SIGNED MESSAGE-----
>Hash: SHA256
>
>So there are then no requirements that rely on rdf:type to directly link
>instances with shapes.
>
>peter
>
>
>On 04/28/2015 05:23 PM, Holger Knublauch wrote:
>> I didn't say that "link instances with shapes" must rely on rdf:type
>> *only*. Of course it could also be a path expression
>> rdf:type/sh:classShape as has been proposed elsewhere. This is certainly
>> a possibility, e.g.
>> 
>> ex:MyClass sh:classShape [ sh:property [ ... ] ; sh:constraint [ ... ] ]
>> .
>> 
>> but I believe the "syntactic sugar" of being able to use classes as
>> shapes directly will be important for the acceptance of this language:
>> 
>> ex:MyClass sh:property [ ... ] ; sh:constraint [ ... ]  .
>> 
>> I was merely reacting to Arthur's statement that my proposal would
>> violate existing agreements. I don't believe it does. And I remain open
>> to adjusting the draft if the WG decides on a level of indirection. My
>> draft clearly marks this issue as unresolved.
>> 
>> Holger
>> 
>> 
>> On 4/29/2015 10:11, Peter F. Patel-Schneider wrote: I do not believe
>>that
>> either of these requirements relies on rdf:type to "link instances with
>> shapes".  They do, of course, require some connection between a class
>>and
>> a shape, but that is different.
>> 
>> peter
>> 
>> 
>> On 04/28/2015 05:05 PM, Holger Knublauch wrote:
>>>>> On 4/29/2015 10:03, Peter F. Patel-Schneider wrote: I would like to
>>>>> see an explanation of which approved requirement relies on rdf:type
>>>>> to "link instances with shapes".
>>>>> 
>>>>>> 
>>>>>>https://www.w3.org/2014/data-shapes/wiki/Requirements#Association_of_
>>>>>>Cla
>ss_
>>
>>>>>> 
>with_Shape
>>>>>> 
>>>>>> 
>>>>>>https://www.w3.org/2014/data-shapes/wiki/Requirements#Selection_by_ty
>>>>>>pe
>>>>>>
>>
>>>>>> 
>HTH Holger
>>>>> 
>>>>> 
>>>>> peter
>>>>> 
>>>>> On 04/28/2015 03:50 PM, Holger Knublauch wrote:
>>>>>>>> Hi Arthur,
>>>>>>>> 
>>>>>>>> On 4/29/2015 5:00, Arthur Ryman wrote:
>>>>>>>>> Holger,
>>>>>>>>> 
>>>>>>>>> Your proposal looks like a revival of a discussion we had
>>>>>>>>> months ago.
>>>>>>>> Yes I remember these long discussions, and I am aware some
>>>>>>>> people are tired of this topic.
>>>>>>>> 
>>>>>>>>> Shapes and Classes are very different things. A Shape is a
>>>>>>>>> set of constraints on a graph. A Class has a set of member
>>>>>>>>> resources that define its extension. The group decided to
>>>>>>>>> keep these concepts separate.
>>>>>>>> Yes, they are conceptually different. But can you point me at
>>>>>>>> a "decision" by the WG that would make my proposal invalid?
>>>>>>>> 
>>>>>>>>> You seem to be repeating the same arguments. What is
>>>>>>>>> different now? Why should the WG reverse its earlier
>>>>>>>>> position?
>>>>>>>> As discussed many times, the WG comprises of different camps.
>>>>>>>> Some people believe Shapes are all they need, others believe
>>>>>>>> that it would be foolish to ignore the existing structures
>>>>>>>> and previous work related to classes and replace it with a
>>>>>>>> parallel universe. We have approved requirements that rely on
>>>>>>>> rdf:type to link instances with shapes. I would appreciate if
>>>>>>>> both sides try to understand each other. My proposal aims at
>>>>>>>> making both view points possible. People who prefer to stay
>>>>>>>> in pure Shapes can use sh:Shape + sh:nodeShape, while others
>>>>>>>> can use rdf:type + rdfs:Class. Yes we could completely
>>>>>>>> separate sh:Shape and rdfs:Class, yet I see more downsides
>>>>>>>> than advantages of such a design. In particular,
>>>>>>>> cross-inheritance of constraints becomes messy if you have
>>>>>>>> rdfs:subClassOf + sh:extends.
>>>>>>>> 
>>>>>>>> Instead of having philosophical discussions, I have opened
>>>>>>>> this thread to try to examine specific proposals, with
>>>>>>>> specific metamodels. If someone has better suggestions, then
>>>>>>>> I would like to read details about them. As this
>>>>>>>> Class-vs-Shapes topic is very important to some people here,
>>>>>>>> I believe we have best chances with a proposal that allows
>>>>>>>> both modeling approaches to be used, and then let the users
>>>>>>>> decide which design they prefer in practice. It's a web-based
>>>>>>>> standard after all.
>>>>>>>> 
>>>>>>>> Thanks Holger
>>>>>>>> 
>>>>>>>> 
>> 
>> 
>-----BEGIN PGP SIGNATURE-----
>Version: GnuPG v2
>
>iQEcBAEBCAAGBQJVQCW2AAoJECjN6+QThfjzNYoIAJ3N/et4tL8z8r3GpP48DkZs
>DDxawIfhHlCrtGp7DVXxyzegG74dN/QayM2r2LKKrJNHuPx5qt7ta5OP3z8/EU81
>82IdToCyT4COQ/OE9iFXceIbEuXV0BAPDYuDXwKgVO25woXCNUNe3s/B2HVUugM8
>dlQRhmzLmd0p+yhCKzpkfhhUDugIJS87O8303TdQpQyc/SkIY1qpKmBelflYT4Ha
>92WGD5s/gwZkliESyckdYgEs/pd4ySnqOru2NrWMkqOs7Wj8czcNpE8jH+a5ocsb
>RoLeCoSJEPyZC5SHBvx2jUQXKUOGfXDQSmH6rVc0puW7WARikmsVY467MlhqLUc=
>=nNrI
>-----END PGP SIGNATURE-----
>

Received on Wednesday, 29 April 2015 00:40:34 UTC