Re: representing deserialized RDF - N3, JSON-LD, RDFa, XML/RDF

Gregg Kellogg
gregg@greggkellogg.net

On Dec 11, 2013, at 7:59 AM, Gregg Kellogg <gregg@greggkellogg.net> wrote:

> On Dec 11, 2013, at 7:47 AM, Ruben Verborgh <ruben.verborgh@ugent.be> wrote:
> 
>>> Turtle is not a standard yet. It's a Candidate Recommendation
>> 
>> Oh yeah of course, I keep treating the CR as a standard. Shouldn't do that.
>> (Still hoping they drop the SPARQL keyword syntax.)
>> 
>>> That's a valid point. In this specific case I would just define a profile
>>> which specifies the structure the document has to adhere to. Since JSON-LD
>>> has a profile media type parameter you can even expose that information on
>>> the wire.
>> 
>> Brilliant, that would be all we need.
>> Could it be such that:
>> - no nesting is allowed (cf. earlier example you posted)
>> - no prefixes are allowed
>> That would give us simple and fast parsers.
>> And those libraries that have full JSON-LD can just use that.
> 
> This is essentially expanded flattened form; it's also what the JSON-LD FromRdf algorithm generates. Note, however, that lists are first-level, and don't use first/rest linking.

Note, however, that there's nothing that would prevent us from using a triple form of expanded JSON-LD, where each triple is in it's own object. Also, rdf:first/rdf:rest can be used in the JSON-LD representation, so use of @list is not necessary; for JSON-LD, toRdf would handle this just fine. Note that Named triples (quads) would require a bit more embedding, e.g.:

{"@id": "http://graph-name", "@graph": {"@id": "http://subject", "http://predicate": [{"@id": "http://object"}]}

Gregg

> Gregg
> 
>> Best,
>> 
>> Ruben
>>

Received on Wednesday, 11 December 2013 16:13:25 UTC