Re: ISSUE-151: Last Call Comment: skos:member definition

Hi,

>> Would a range statement be considered as too big a change for the spec?
>> I would live very comfortably with a range axiom for skos:member, as 
>> the naming of the property is very ambiguous. And we don't think it 
>> will be used with other kind of objects, do we?
>
>
> I think ading a range constraint should be fine, as we're not really 
> changing the design, only making it more precise. Can you make a 
> concrete proposal for the range constraint you'd like to add to Ref? 
> We would have to weigh carefully whether it does not break some use 
> cases, however.

The constraint would be
[
The range of skos:member is the union of skos:Concept and skos:Collection
]

which we can add in the OWL ontology for SKOS as
[
skos:member rdfs:range [ owl:intersectionOf (ex:bicycles ex:repairing) ] .
]
(if I'm not mistaken with Turtle syntax)

About Alistair's comment:
> But why include the range statement? Who needs it?

I agree this change does not appear at first sight very crucial. It could even harm flexibility, as you said in your first answer. I have three answers to that:
1. coherence of our specs. It would make ou formal spec more in line with the text (why say that collections are collections of concepts if we secretely allow for other individuals? I perfectly understand Erik's being puzzled here).

2. coherence of SKOS data. In fact I would prefer to have the collections constrained than finding later that some users have created instances of collections that include, say, instances of skos:Concept and instances of skos-xl:Label. 

3. Understandability of collections. The only use case we have for collections is thesaurus arrays, and these are just meant to include thesaurus terms, aren't they? Our past experience shows that collections are difficult for users to understand. If we loose the link to thesaurus arrays by saying that collections are grouping of unconstrained things, it would make even more difficult to understand when using collections or not. 

Cheers,

Antoine


>
> Guus
>
>>
>> Antoine
>>
>>> Here is a draft response to Erik on ISSUE-151, comments welcome.
>>>
>>> --- begin draft message ---
>>>
>>> Dear Erik,
>>>
>>> Thank you for your helpful comments. In response to the comment below:
>>>
>>> On Wed, Oct 01, 2008 at 09:20:03PM +0000, SWD Issue Tracker wrote:
>>>  
>>>> ISSUE-151: Last Call Comment: skos:member definition
>>>> http://www.w3.org/2006/07/SWD/track/issues/151
>>>>
>>>> Raised by:  Everyone
>>>> On product: All
>>>>
>>>> Raised by Erik Hennum in [1]:
>>>>
>>>> """
>>>> Should the specification define skos:member as having a range of
>>>> skos:Concept or skos:Collection? Should skos:member have an inverse
>>>> skos:isMemberOf property?
>>>> """
>>>>     
>>>
>>> We have not encountered any requirements to specify the range of
>>> skos:member. We propose to make no change to the current draft,
>>> leaving the range unspecified, allowing greater flexibility in the use
>>> of the SKOS collections framework, for example with third party
>>> extensions. Can you live with this?
>>>
>>> Similarly we have not encountered a requirement for an inverse of
>>> skos:member. We propose to make no change, can you live with this?
>>>
>>> Kind regards,
>>>
>>> Alistair
>>> Sean
>>>
>>> [1] http://lists.w3.org/Archives/Public/public-swd-wg/2008Jun/0103.html
>>>
>>>   
>>
>>

Received on Friday, 24 October 2008 08:28:27 UTC