W3C home > Mailing lists > Public > public-rdf-comments@w3.org > May 2012

Re: JSON-LD Syntax request for FPWD via RDF WG

From: Gregg Kellogg <gregg@greggkellogg.net>
Date: Tue, 22 May 2012 15:00:12 -0400
To: Richard Cyganiak <richard.cyganiak@deri.org>
CC: Markus Lanthaler <markus.lanthaler@gmx.net>, "public-rdf-comments@w3.org" <public-rdf-comments@w3.org>
Message-ID: <70C02FDC-7D40-4DDE-BBD9-163428B28236@greggkellogg.net>
On May 22, 2012, at 10:42 AM, "Richard Cyganiak" <richard.cyganiak@deri.org> wrote:

> Hi Gregg,
> 
> On 22 May 2012, at 17:54, Gregg Kellogg wrote:
>> As Markus noted, at this point, it's only the Syntax spec that we're submitting. The API spec could potentially be done in a different group, when the time comes.
> 
> I feel about this like Ivan. There is a line about producing a JSON syntax for RDF in the RDF-WG charter. The JSON-LD syntax spec, on its own, is not sufficient to be called a JSON syntax for RDF. I'm sure you see the problem with saying that the missing piece (to- and from-RDF algorithms) will be delivered by some other WG at some future date.

If moving the algorithm here helps, then we should probably do that. In this case, it should probably be divorced from the WebIDL constraints of the API document, and can directly represent RDF Concepts.

>> The use of terms such as _Statement_ closely follows the RDF Interfaces spec [3] (Triple renamed to Statement), which has been dormant.
> 
> If they were not dormant, I'd have a word with them. Still, this document is quite a bit closer to RDF Concepts than the JSON-LD API document.
> 
>> I think it's reasonable that these terms echo definitions in RDF Concepts, but note that a Statement may be either a triple or a quad; triple seems too narrow for this. I don't believe the concepts doc discusses triples with a context.
> 
> It doesn't yet, but it will. (This is the “RDF dataset”/“named graphs” stuff consumes 75% of this WG's traffic.) Opinions differ on whether we want quads. Personally I'm opposed, and prefer a view where we have multiple graphs, each made up of triples, that are associated with a IRI (graph name), like SPARQL's RDF datasets. This is what's currently in the RDF 1.1 Concepts Editor's Draft, but it will still go through some iterations:
> http://dvcs.w3.org/hg/rdf/raw-file/default/rdf-concepts/index.html#section-multigraph


The algorithmic issues for representing empty names graphs are pretty much the same as for TriG. The syntactic representation is not an issue in either syntax. Perhaps a type mapping in the default graph that asserts that the graph name is some kind of rdf:Graph; it's not specifically a JSON-LD issue.

> Are there specific use cases that you have considered in order to design the quads/context feature of JSON-LD? We might want to add them to our list of use cases for the multigraph stuff.

Yhe two main use cases were PaySwarm graph signing and Wikidata provenance representation. I think these are both items in your wiki now.

>> For the context of the RDF WG, we could create a separate document describing the normative requirements for RDF transformations; we intentionally kept the discussion of RDF to a minimum in the Syntax document.
> 
> Just speaking for myself, and without having considered the issue deeply, I'd prefer having everything that is required to convert between JSON-LD and an RDF graph in a single document.

Out intended audience was Linked Data developers, for whom RDF is not always appreciated. Keeping these details in the "Advanced Concepts" section, or a normative appendix will probably suffice. I'll let other CG members chime in.
 
>> JSON-LD can pretty fully represent everything that can be represented in TriG, with the exception of lists containing other lists.
> 
> Cool.
> 
>>>> 2. Examples would be great.
>>> 
>>> There are a couple of example in the syntax spec [2], don't know if you
>>> already saw them.
>> 
>> A good source of examples is the Test Suite [4], [5]. We should probably create links from the test suite to each individual test and result, to make them easier to access.
> 
> JSON-LD is all about linked data, right? So I'd expect to see hyperlinks (that is, full URLs) in [4] and [5].

Similar to RDFa or Turtle, JSON-LD supports relative IRIs. If you resolve those references relative to the manifest location, you'll get the full IRI. You can turn it into RDF using my distiller (at http://rdf.greggkellogg.net/distiller. You could also use that to turn arbitrary RDF into JSON-LD.

Gregg

>> The fact that manifests are all represented using JSON should make this a fairly easy thing to do within the HTML page itself, perhaps using the <script type="application/ld+json"> similar to that used in the Turtle spec.
> 
> Sounds useful.
> 
> Thanks,
> Richard
> 
> 
> 
>> 
>> Gregg
>> 
>>>> 3. Is it possible to serialize an RDF graph into a "pretty" JSON-LD
>>>> document using a context? I presume the answer is "yes" and involves
>>>> Compaction of the basic serialized output.
>>> 
>>> Yes, exactly either by compacting or by framing.
>>> 
>>> 
>>> [1] https://github.com/json-ld/json-ld.org/issues/125

>>> [2] http://json-ld.org/spec/latest/json-ld-syntax/#markup-examples

>> [3] http://www.w3.org/TR/2011/WD-rdf-interfaces-20110510/

>> [4] http://json-ld.org/test-suite/tests/fromRdf-manifest.jsonld

>> [5] http://json-ld.org/test-suite/tests/toRdf-manifest.jsonld

>> 
>>> --
>>> Markus Lanthaler
>>> @markuslanthaler
>>> 
>>> 
>>> 
>>> 
>>> 
>> 
>> 
> 
Received on Tuesday, 22 May 2012 19:01:21 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Tuesday, 22 May 2012 19:01:21 GMT