Re: [JSON] RDF collections and JSON arrays

On Mar 25, 2011, at 14:41 , Pat Hayes wrote:

> 
> On Mar 25, 2011, at 4:12 AM, Nathan wrote:
> 
>> Steve Harris wrote:
>>> I'm also not keen on [[ ]] for RDF collections, but don't see a sensible alternative.
>> 
>> The main alternatives are to simply not include support for either multiple values or RDF Collections.
> 
> I dont think not supporting RDF collections is an option. There is lots of RDF out there that has a few pieces of OWL in it. OWL syntax in OWL/RDF uses RDF collections. If we don't support collections well enough to handle this, all the RDF<->JSON will break whenever it hits a piece of OWL. I think this would be a disaster, and would cause enough grief to make the JSON unusable in the wild.

Pat,

much as I would like to have collections folded in somehow, I am not sure that statement is right. The discussion is a bit similar to what we had on the RDFa WG on whether RDFa should have some syntactic sugar for collections (and the decision is: no).

I agree that collections are heavily used in OWL. However... my _feeling_ is that JSON will be used, primarily, for instance data. Ie, an RDF environment may of course have a reference to an OWL ontology, but the ontology itself, can be encoded in some other formats. That is because, again I believe so, applications that would really make use of the inference possibilities of OWL would be in some sort of a multi-parsing environment anyway, so they will not care.

Put it another way: I do not think that many ontologies will be encoded in JSON. They will happily go on using turtle or, God forbid, RDF/XML, produced by tools like Protégé. We may not want to optimize on those.

Ivan


> 
> Pat
> 
>> 
>> [[ ]] is unexpected for JSON users, and I fear that in reality people will simply think we are mad, after all the context under which I proposed the [[ ]] syntax was for a machine optimized RDF+N3 in JSON, and not what we appear to be looking at creating now (what I'd refer to loosely as Data Objects).
>> 
>> Looking at this practically, most of our target market do not have the concept of multiple values, nor do they have support for multiple values in their stacks. They do however work with arrays and collections intensively.
>> 
>> I'd actually propose that we drop multiple value support and keep the [] notation for RDF collections.
>> 
>> To illustrate, if somebody has a property with multiple values for the label, they will typically write:
>> 
>> "labels": ["Foo", "Bar"],
>> 
>> Notice that's plural "labels" not singular "label", my way of language they are saying "this is a collection of labels", and that is the most common usage - indeed it's practically about the only option many developers have in their non-RDF tech stacks.
>> 
>> This approach of only supporting RDF Collections (producing JS Arrays) makes sense to people. [[ ]] syntax and "multiple values" do not.
>> 
>> On the RDF side, surely we can create some properties if we need to? rdf:labels -> a collection of rdf:label relating to the subject.
>> 
>> Best,
>> 
>> Nathan
>> 
> 
> ------------------------------------------------------------
> IHMC                                     (850)434 8903 or (650)494 3973   
> 40 South Alcaniz St.           (850)202 4416   office
> Pensacola                            (850)202 4440   fax
> FL 32502                              (850)291 0667   mobile
> phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes
> 
> 
> 
> 
> 


----
Ivan Herman, W3C Semantic Web Activity Lead
Home: http://www.w3.org/People/Ivan/
mobile: +31-641044153
PGP Key: http://www.ivan-herman.net/pgpkey.html
FOAF: http://www.ivan-herman.net/foaf.rdf