Re: [JSON] Survey for design requirements

Ivan Herman wrote:
> On Mar 7, 2011, at 13:37 , Nathan wrote:
> 
>> Michael Hausenblas wrote:
>>>> What I am trying to find is the sweet spot that would create a bridge between them and those Web Application Developers who do not know RDF or, worse, who are averse to RDF because they see it as too complex, researchy, etc (I think we all met these various memes around that). Doesn't that mean that to please those a JSON/RDF must be damn simple (even if it does not cover the whole of RDF), and maybe also include shorthands to make it even simpler (eg, default subjects to see only property value pairs, etc). At first glance this hints at something like the 's', 'p', 'o' format of Andy with some defaults here and there...
>>>>
>>>> Put it another way, I am not sure our goal is to build a JSON serialization that would compete with Turtle. Only with N-Triples...
>>> +1
>>> I think you have a very valid point, here. I came up with a proposal, called JTriples [1] - not pretty, still some open issues (re literals) but might serve as a basis for exploration.
>> I hate to say this, but if the problem is web developers who do not know RDF or, worse, who are averse to RDF because they see it as too complex, then surely the last thing we want to do is give them raw RDF triples in almost their purest form?
>>
> 
> Well, it is difficult to say. I cannot really judge it and I am a bit afraid none of us can. But... What are RDF triples in their purest form? Property-value pairs with an additional thing called subject. If we also have some mechanism to define a default subject or common subjects, then that is still very simple. Maybe that is _all_ that most of the Web Application programmers want to know about! Anything else is extra complication...

Yes, I agree :) hence why mentioning,

{s: subject1, p: property1, o: value1 }
{s: subject1, p: property2, o: value2 }

vs

{
   id: subject,
   property1: value1,
   property2: value2,
}

> There are of course issues around URI-s vs. Literals: well, we had some fierce discussions at the RDFa Working Group on whether RDFa would include some automatism that says that if a literal in an object position can be interpreted like a URI, then we should generate a URI Resource and not a Literal. There are millions of details there but, again, a simple minded mechanism like that would cover 80% of those developers' use cases.
> 
> Again, I am not 100% sure what this means but what I am trying to say is that, for the JSON serialization, maybe we should try to look at the issue with a totally different mindset...

agree

Received on Monday, 7 March 2011 12:55:23 UTC