W3C home > Mailing lists > Public > public-rdf-comments@w3.org > June 2013

Re: Official response to RDF-ISSUE-132: JSON-LD/RDF Alignment

From: Kingsley Idehen <kidehen@openlinksw.com>
Date: Mon, 10 Jun 2013 14:36:54 -0400
Message-ID: <51B61CC6.7080503@openlinksw.com>
To: public-rdf-comments@w3.org
On 6/10/13 1:46 PM, Pat Hayes wrote:
> On Jun 8, 2013, at 3:39 PM, Kingsley Idehen wrote:
>> On 6/8/13 2:45 PM, David Booth wrote:
>>> On 06/08/2013 02:30 PM, Markus Lanthaler wrote:
>>>> On Saturday, June 08, 2013 5:13 PM, Sven R. Kunze wrote:
>>> [ . . . ]
>>>>> Would adding your CLEAR statement ("JSON-LD is a concrete syntax of
>>>>> RDF.") to the part "relationship" or even in the intro as the first
>>>>> line as e.g. turtle does do any harm?
>>>> You didn't read the intro, but already there we say
>>>>     Developers that require any of the facilities listed above or
>>>>     need to serialize an RDF graph or dataset [RDF11-CONCEPTS] in
>>>>     a JSON-based syntax will find JSON-LD of interest.
>>>> IMHO that's crystal clear.
>>> But clarity needs to be assessed in the eyes of the *readers* -- not the authors.  Obviously it is *not* clear to readers, as you've heard substantially similar comments on this point from both Sven and from me.  Telling readers that they "will find JSON-LD of interest" is nowhere near as clear as explicitly saying that "JSON-LD is a concrete syntax of RDF" or "JSON-LD is a serialization format for RDF".
>>> David
>> JSON-LD can be used to produce Linked Data. It can leverage RDF as a mechanism for increasing the semantic fidelity of the Relations in the Linked Data produced.
>> RDF cannot be all things to all scenarios. That's the darn problem with the narratives around RDF that have utterly obscured its comprehension and appreciation over the years.
>> RDF is extremely useful. It can stand on its own merits. It doesn't need to add Linked Data to that bucket too.
> That is not the point. The point is that JSON-LD is *obviously* modelled on RDF, is *designed* to be as close to the RDF data model as possible, but its documentation *deliberately* obscures this fact , and moreover it is clear from reading the email archives that this is a conscious decision, intended to mislead readers.

JSON-LD wasn't about misleading the readers. That characterization is 
really unfair. I was an active participant in the discussions about 
Linked Data. In all cases, I make one fundamental point:

If you are going to refer to TimBL's meme (revised edition at added RDF 
and SPARQL to the mix) as the Linked Data dictum then by implication you 
have RDF in the mix too. If on the other hand, you are looking at Linked 
Data as concept where RDF and SPARQL are implementation details, then 
also do so with clarity.

JSON-LD eventually settled for a partial variant of what I outlined above.

>   Such behaviour is IMO not appropriate when writing standards.

Yes, if that is true. And by the way, thanks a zillion for chiming in, 
at least I am now hopeful that some nuances can be ironed out once and 
for all re. Linked Data and RDF.

>   I recognize that it is motivated by a desire to connect with a sizeable body of people like yourself who would be irrationally "put off" by seeing the three letters R D F anywhere in the spec, or possibly even by the mere word "semantic", but IMO that does not justify writing a specification document which sets out to deliberately obscure, rather than document, an important technical connection between specifications. It is basically dishonest, is the main problem I have with the JSON-LD spec. Specs should not be thought of as advertising or propaganda.

I actually coined the phrase R-D-F reflux. I didn't have myself in mind 
that I coined the phrase. It was about a profile out there that 
instinctively reacts to R-D-F. At this point, I can only assume you 
haven't had time to comb through this massive debate to clearly access 
my position. My position isn't divergent from yours, since I've spent 
years trying to simplify and articulate nuances you've actually also 
raised over the years about HTTP URI ambiguity and its implications etc.

>> JSON-LD is also very useful, it too can stand on its own merits as a mechanism for producing Linked Data that's palatable to JSON developers while also enabling exploitation of RDF's virtues.
> Your point here presumes that JSON-LD is different from RDF, so provides an alternative.

My point assumes that you can use JSON-LD to produce RDF model based 
structured data representation. It means, that a Linked Data URI can 
resolve to a description document comprised of RDF model based content 
encoded using JSON-LD.

>   Which would be a very good point if it were true, but it (basically) isn't.

Bearing in mind the comment above, how isn't that true?

>   (Actually it is when you get down to picky details like using bnodes as properties and literal formats, but these mismatches between JSON-LD and RDF are widely seen as a *problem*. Why would that be, if these are distinct alternatives?) JSON-LD is basically a JSON syntax for RDF (with a few sore rubbing spots). Which is a great thing, and should be encouraged, but let us at least be honest about what it actually is.

/"JSON-LD is basically a JSON syntax for RDF (with a few sore rubbing 
spots). Which is a great thing, and should be encouraged"/

In the excerpt above you are expressing my basic point while somehow 
trying to assume I am taking an alternative position. Seriously now, 
what's the difference between what you are saying and the position I am 
actually taking?

JSON-LD is really about Linked Data first, and then RDF second. In 
reality though, JSON-LD is much more about how to use JSON to produce an 
entity relationship model based description graph where URIs conform to 
the original principles outlined in TimBL's meme i.e., use of HTTP URIs 
to enhance structured data representation by making it web-like (or webby).

Linked Data isn't a subset of RDF. It isn't something you can only 
produce based on the RDF spec, far from it.

>> The Web works because its components are loosely coupled.
>> RDF has struggled over the years because positions like yours are completely antithetical to what makes the Web work.
>> The Web was semantically interlinked from day one, and there was no RDF on day one.
> The Web was linked from day one, but it was not *semantically* linked until there was some semantics, and the RDF data model provided that. It's a very minimal semantics, but it's better than nothing.
>> Without Links there is no Web. On the Web a Link denotes a Relation.
> Really? Who says? I have yet to see any spec which provides a semantics for web links. (And in any case, a relation between what?)

A link (as you now) denotes a Relation that has a subject, predicate, 
and an object, in RDF parlance. A link (as far as I know) associates one 
thing with another. You have:

thing1-- link --> thing2.

You can denote each item in the statement above using literals or 
references. In RDF (at the current time) thing1 and link are denoted 
using IRIs. You can denote thing2 using IRIs or literals.  The item 
denoted by "link" above is the relation. The thing that has domain and 
range properties etc..

>> When Links denote Relations you have a Semantic Web,
> So the Web has been Semantic since 1991? How about that, all that time and nobody knew!

I think quite a few people knew that since they understand what links 
are, long before the World Wide Web emerged. The Web didn't invent 
links, it just amplified their power via hypermedia resources woven 
together using HTTP URIs. The meaning of the links is what's been 
evolving and that's where RDF comes into play since it enables one 
explicitly express the semantics of a link i.e., be specific about what 
kind of relationship associates the subject and objects of a relation.


1. http://dbpedia.org/resource/Linked_data -- I had that returning 
JSON-LD once the basic definition of the effort was clear to me (yes, 
while everyone was arguing about formats as per usual, I had it 
incorporated into our engine that drives DBpedia

2. http://bit.ly/15lpf1K -- Linked Data URI verifier report (note: it 
support JSON-LD inline with what you outlined and what I just explained 
to you was my position too)

Re. item #2 the report even states (and btw I also had this verifier 
tweaked so that it's RDF/XML specificity was removed):
**Conclusions on the type of the resources**:*

<http://dbpedia.org/data/Linked_data.jsld> denotes a Web document 
bearing JSON-LD content
<http://dbpedia.org/resource/Linked_data> denotes an Entity described by 
the JSON-LD content.

I am sure you can now see what my intentions are, especially as they are 
clearly expressed in any Linked Data technology where I have actual 
technical involvement.

I use RDF extensively, appreciate it immensely, and I get really ticked 
off when I see people (maybe well intended most of the time) mucking 
things up on the narrative and evangelism fronts. Time and time again, 
there is an all or nothing mindset that simply plays into the hands of 
its detractors.

I also use Linked Data extensively, I understand what it is (profoundly) 
and I also get really ticked off when I see people (maybe well intended 
most of the time) mucking things up. If RDF is an all or nothing 
proposition, then do we have to drag Linked Data down that same rat-hole?

Here's my simple question to you, bearing mind my response to your 
misunderstanding of my position and intentions. What's so wrong with 
"RDF based Linked Data" as a clear way of establishing the fact that 
when it comes to Linked Data the role of RDF is that of a means of 
production rather than the concept itself?

BTW -- I've never heard you (in anyway) conflate RDF and Linked Data. My 
frustration and ire is aimed squarely at those who still don't 
understand that by conflating Linked Data and RDF you end up with a 
disservice to both, along comprehension and adoption dimensions.

> Pat
>> even if the granularity of the semantics are coarse rather than fine-grained to the perceptive prowess of humans and/or machines.
>> The RDF narrative doesn't have to be so warped to the point of being ridiculed at every turn by folks who know far less, but have much deeper pockets and alternative ideas about what the Web is all about.
>> If TimBL meant Linked Data to be a subset of RDF, rather than something you can produce using RDF, don't you think he would know exactly how to make that point?  He wouldn't need you or I to tell him how to pen such a meme, note, or set of guidelines.
>> -- 
>> Regards,
>> Kingsley Idehen	
>> Founder & CEO
>> OpenLink Software
>> Company Web: http://www.openlinksw.com
>> Personal Weblog: http://www.openlinksw.com/blog/~kidehen
>> Twitter/Identi.ca handle: @kidehen
>> Google+ Profile: https://plus.google.com/112399767740508618350/about
>> LinkedIn Profile: http://www.linkedin.com/in/kidehen
> ------------------------------------------------------------
> IHMC                                     (850)434 8903 or (650)494 3973
> 40 South Alcaniz St.           (850)202 4416   office
> Pensacola                            (850)202 4440   fax
> FL 32502                              (850)291 0667   mobile
> phayesAT-SIGNihmc.us       http://www.ihmc.us/users/phayes



Kingsley Idehen	
Founder & CEO
OpenLink Software
Company Web: http://www.openlinksw.com
Personal Weblog: http://www.openlinksw.com/blog/~kidehen
Twitter/Identi.ca handle: @kidehen
Google+ Profile: https://plus.google.com/112399767740508618350/about
LinkedIn Profile: http://www.linkedin.com/in/kidehen

Received on Monday, 10 June 2013 18:37:22 UTC

This archive was generated by hypermail 2.4.0 : Friday, 17 January 2020 16:59:34 UTC