Re: JSON-LD Telecon Minutes for 2011-07-04

On 07/12/2011 02:50 PM, Kingsley Idehen wrote:
>> The question is, do unnamed nodes have a place in Linked Data and in
>> JSON-LD? I think they do; anyone else?
>>
> No, not if you want to deliver a simple (lightweight) mechanism for
> constructing Linked Data graphs using JSON-LD. The moment you introduce
> anonymous references (blank nodes) you've introduced an RDFism. Net
> effect, inheritance kicks in and we are back to square one re. inability
> to deliver a simple JSON based mechanism, with high uptake potential
> re.,  Linked Data graphs.

I think we're skirting very close to letting theoretical purity get in
the way of getting real work done. If we don't allow unlabeled nodes (or
whatever you want to call them), then this is illegal in JSON-LD:

{
   "name": "Kingsley Idehen"
}

If that is an illegal statement in JSON-LD, then JSON-LD is certainly
not a sub-set of JSON, but a newer, more limited form of JSON - and that
is when just about every JSON developer will look away from JSON-LD and
keep looking for something that will work for them in the real-world.

If we want to get pedantic, an unlabeled node is not an RDFism, it's a
purely mathematical concept:

http://en.wikipedia.org/wiki/Labeled_graph

> In the early days of Linked Data we agreed to discourage blank nodes for
> the very reasons outlined above. Basically, there was a concerted effort
> to focus on uptake and bootstrap.

I appreciate what you are doing, Kingsley - which is to try and steer us
toward the simplest solution. :)

I'm coming at it from the direction of a company that doesn't want to
have to generate a IRI for every piece of data that they want to talk
about. In many cases, associating the 'unlabeled' data with some data
that /has/ an IRI is exactly what we want.

So, if your goal is to focus on uptake and bootstrap - it will have the
opposite effect on our company. It will generate /more work/ for us
because we will have to teach all of our engineers how to create lasting
identifiers for every way that we group our data online. It will
generate /more work/ for us to create some sort of distributed algorithm
for generating digital signature IDs (which we'll need to do if we don't
support unlabeled nodes in JSON-LD). It will generate /more/ work for us
when one of our developers inevitably gets one of the many ID generation
algorithms wrong. :/

Rather than generate all of this work - we could just support unlabeled
nodes. It's not an RDFism - it's a math-ism. By attempting to make this
part "simpler" by not supporting unlabeled graphs, we harm uptake.

Try telling someone that has been living and breathing JavaScript and
JSON for the past 5 years that they can no longer do this:

{
   "name": "Kingsley Idehen"
}

I doubt that they will be very happy with that restriction. :)

-- manu

-- 
Manu Sporny (skype: msporny, twitter: manusporny)
President/CEO - Digital Bazaar, Inc.
blog: PaySwarm Developer Tools and Demo Released
http://digitalbazaar.com/2011/05/05/payswarm-sandbox/

Received on Tuesday, 19 July 2011 01:00:55 UTC