Re: A modest attempt to re-open ISSUE-20

On 3/15/13 12:49 PM, Erik Wilde wrote:
> hello kingsley.
>
>> 1. RDF (a framework comprised of: model, syntax, syntax notation, and
>> serialization formats) isn't the same thing as Linked Data (which is an
>> application of RDF for creating graphs where URIs based hyperlinks have
>> specific behavioral expectation or functionality)
>
> i think i see where you want to go with this, but at the linked data 
> level, the problem i have with calling this "behavioral expectation" 
> is that it's a read-only model (after all, there;'s a reason why it's 
> called "linked data" and not "linked services"), which is a rather 
> limited behavioral model.

No, Linked Data isn't read-only. It is just about the use of hyperlinks 
to enhance the construction of entity relationship model based graphs. 
Put differently, its webby structured data that fully exploits Web 
architecture.

Linked Data is all about a principled approach to publishing structured 
data on the Web. That doesn't mean its read-only though.

Linked Data URIs are just conduction mechanisms for CRUD operations on 
Data.


>
>> As proposed some time ago, maybe we need a content-type for RDF based
>> Linked Data. This would enable RDF heavy and RDF Lite solutions to be
>> less confused about payloads exchanged over HTTP (and any other protocol
>> in the future).
>
> i think there's two ways to go here. one way is to do what you say and 
> have a media type that exposes linked data. however, rather than doing 
> this, i think it might be better to go the "hyperRDF" way and have a 
> model that's capable of expressing any kind of hyperlinked resources, 
> and not just read-only content.

See my comments above. The only issues here is comprehension of how 
Linked Data URIs align with the long understood semiotic triangle [1].
>
> notice, though, that the other popular web-level data models (XML and 
> JSON) have never done this.

JSON-LD has a content-type for understanding how it enables exploitation 
of RDF model based Linked Data.

> they have no model for hyperlinks, but instead rely on more specific 
> models to add hypermedia semantics.

No, they enable mapping to  RDF (as the data model) and exploitation of 
RDF model based Linked Data. This is factored into the content-type 
associated with specific type of JSON.

>
> thus, the second way to go is to have specific media types for models 
> which expose hypermedia controls. this is what happens in the XML and 
> JSON worlds (proposals to create "hyperXML" and "hyperJSON" so far 
> have been met with reluctance), but is not what the LDP wants to do.

As per my suggestion, the way out of this quagmire is to simply have an 
RDF Lite based Linked Data content-type. This loosens the coupling in 
ways that ultimately keep all the parties happy. Conflation hasn't 
worked to date (talking 13+ years now) and it won't start to 
miraculously work anytime in the future.
>
> either way, it definitely would be good to expose LDP as a hypermedia 
> media type.

I don't know about LDP, but Linked Data will certainly benefit from such 
a thing.

> and it may actually set an example for other linked data services, and 
> therefore it's great to have this discussion.

Yes, it's very important. The issue of conflation is causing too many 
distractions :-)

Links:

1. http://twitpic.com/cbk8ul -- Linked Data URIs and the Semiotic 
Triangle .

>
> 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