Re: LDP interfaces in Java (based on Jena and JAX-RS)

On 8/6/12 8:09 AM, Erik.Wilde@emc.com wrote:
> hello kingsley.
>
> On 2012-08-06 13:42 , "Kingsley Idehen" <kidehen@openlinksw.com> wrote:
>> RDF is simply an option re. Linked Data. What's important is the
>> behavior of URIs, not a specific family of data representation syntaxes
>> and serialization formats.
> that's one possible view of what "Linked Data" means and one i had a while
> ago. but it seems that the majority of the linked data community has a
> more constrained understanding of what it means, and when i used the term
> to just mean what it says, literally, i.e. "data that is linked", it
> seemed the linked date community largely objected to that and claimed that
> http://www.w3.org/DesignIssues/LinkedData.html mandated the use RDF.

If you take the literal route they'll counter with TimBL's design issues 
note [1]. Thus, that isn't the route I would advice when attempting to 
bring clarity to this most important of matters.

Some history:

When TimBL first crafted his Linked Data meme it made no mention of RDF, 
and for good reason. Basically, his fundamental vision of the Web [2] 
was never one constrained by any data representation or serialization 
format.

The W3C (like many others) sought to ride the Linked Data meme's wave by 
inserting RDF into the meme. Now, there's nothing wrong with that per 
se., if the net effect protects the meme's essence and accelerates 
comprehension and adoption. Unfortunately, due to the history 
surrounding RDF (technical and political) the insertion of RDF 
(explicitly) into the meme has lead to nothing but confusion.


>
> while i am still of the opinion that it would be more useful to talk about
> linked data as a concept of "how to properly make data available in the
> framework of web architecture" (as you suggest), i have stopped using the
> term "Linked Data" in that more general sense, mostly to avoid making
> people unhappy.

Linked Data, as even espoused by TimBL's meme, is fundamentally about 
structured data representation combined with the intrinsic prowess of 
de-referencable URIs. Point #3 in the revised meme is a problem because 
it opens the door for problematic RDF conflation. IMHO., the revisions 
are trying to articulate the fact that its highly recommended that you 
use W3C standards (such as RDF, SPARQL) for the requisite web document 
content (the description graphs)  and indirection that facilitates this 
specific kind of URI de-reference .
>
> of course, in the end everybody can use the term as they see fit, because
> it is not a registered trademark, but realistically speaking given the
> current landscape, "Linked Data" implies RDF.

It implies EAV/CR or RDF or any other 3-tuple model for structured data 
representation where:

1. URIs denote description subjects
2. URIs denote attributes, properties, or predicates
3. URIs are option values of attributes, properties, or predicates based 
on 1&2 above
4. URIs resolve to useful information (data in context) delivered by web 
documents identified by a denotative URI.

>
> you say "What's important is the behavior of URIs, not a specific family
> of data representation syntaxes and serialization formats", and that's
> much more REST than it is RDF.

Yes, because RDF has little to do with REST while Linked Data is 
actually all about REST. Put differently, Linked Data enables RESTful 
exploitation of RDF.

> REST asks for some other constraints, but
> one core constraint is that resource are identified by URI, that URIs can
> be used for interactions using a variety of schemes, and that a RESTful
> system is driven by clients following typed links that they find in
> representations.

Yes!

And a Linked Data resource is an extremely powerful kind of 
self-describing resource that meshes most naturally with REST.

>
> the main issue with RDF in the context of REST is that it has no notion of
> links; all it says (in the Linked Data flavor) is that you can GET
> everything that has an identity.

Modulo Linked Data pattern it has not notion of de-reference. Without 
de-reference there are no resources to interact with RESTFully or 
otherwise.

> this conflation of identification and
> interaction is what makes it hard to envision RDF being RESTful,

RDF doesn't address de-reference becuase it doesn't care about URIs bar 
their function as a denotation mechanism.

>   because
> REST asks of all representations to have hypermedia semantics.

It requires de-reference and self-describing resources.


Some good news re. RDF clarity, the current workgroup is finishing a set 
of documents that bring a lot more clarity to RDF. This work will make 
it much easier to decouple RDF and Linked Data without any adverse 
effects to either.

To conclude, I encourage you to look at things this way:

1. entity-attribute-value plus classes & relationship (EAV/CR) is an 
established model for structured data representation
2. add URIs to EAV/CR and you get RDF
3. add de-referencable URIs to RDF and you have one route to Linked 
Data, consistent with the principles outlined in TimBL's meme.

As Dave Beckett (@dajobe) once said: Linked Data is about webby data. I 
would embelish that a little and say its about webby structured data :)


Links:

1. http://www.w3.org/DesignIssues/LinkedData.html -- TimBL's design 
issues note
2. http://www.w3.org/DesignIssues/diagrams/history/proposal-fig1.gif -- 
original web design illustration (it was always about Linked Data via 
*descriptor* resources)
3. http://lab.linkeddata.deri.ie/2010/star-scheme-by-example/ -- 5-star 
Linked Data (note: there's no RDF specificity in this note).
>
> cheers,
>
> dret.
>
>
>
>


-- 

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