Re: Named Graphs / Layers / Surfaces / Doohickies added to JSON-LD

On 30/04/12 23:46, Sandro Hawke wrote:
> On Mon, 2012-04-30 at 15:56 -0400, Manu Sporny wrote:
>> Our intent when we started JSON-LD was to leave named graphs out of the
>> spec. We were waiting on this group to finalize the modeling aspects of
>> named graphs because we didn't want to do something counter to what this
>> group was going to propose.
>>
>> We had been pushing back on requests to add named graphs to JSON-LD for
>> quite some time and finally had to give in at the end because we had to
>> understand how named graphs might affect the syntax in the future. We
>> didn't want to paint ourselves into a corner. In the end, it was a
>> fairly benign thing to add (from a syntax perspective), so we just went
>> ahead and did it.
>>
>> Keep in mind that we go out of our way to not mention how advanced
>> concepts like sharing bnode identifiers between named graphs works (or
>> doesn't work). In other words, we specified the syntax for naming
>> graphs, but have not really addressed any of the range issues since
>> we're waiting on the RDF WG to propose something.
>>
>> Section 4.9 introduces the concept of a Named Graph in JSON-LD:
>>
>> http://json-ld.org/spec/latest/json-ld-syntax/
>
> I'm not quite sure I understand the syntax, but if my guesses are right,
> it seems like a reasonable design.   Excellent.   Glad to hear JSON-LD
> is ready for broader attention.
>
> Back on the subject of layers, I note that your use of the term "graph"
> in this spec is, I think, evidence of the problem that's been causing us
> such problems, and which I think the term "layers" solves:
>
>          This example says that there is a linked data graph identified
>          by http://example.org/linked-data-graph which is composed of the
>          statements about Manu and Gregg and a reference to another IRI,
>          which could make statements about Markus.
>
> This text uses the word "graph" to mean g-box (now "layer", for me), not
> to mean RDF Graph.   At least, I'm 99% sure it does.  It makes almost no
> sense to talk about an RDF Graph that way, giving it a human-readable
> HTTP URL.

It may be a manifestation of the fact Pat pointed out and has come up 
several times, if the container is unchanging, using the name of the 
container as a way to talk about the contents (the mapping being 1-1, 
time invariant) is something we all do.

 Andy

Received on Tuesday, 1 May 2012 11:38:49 UTC