Re: JSON-LD terminology

On 8/31/12 7:51 AM, Andy Seaborne wrote:
>
>
> On 31/08/12 12:29, Kingsley Idehen wrote:
>> It claims any RDF document is a Linked Data document. That simply isn't
>> accurate. The RDF spec makes no claims about IRIs being:
>>
>> 1. de-referencable
>> 2. http: scheme based.
>
> Let's talk about what the document actually says in section 3.1 and 
> not make comments about political opinion and what one "ideal world" 
> might look like.
>
> The document says:
>
> [[
> 7. IRIs used within a linked data graph SHOULD be dereferenceable to a 
> Linked Data document describing the resource denoted by that IRI.
> ]]
>
> SHOULD, not MUST.
>
> In reality, not all IRIs are going to be dereferencable.
>
> The condition should be "potentially dereferencable" because the 
> burden is far too high otherwise.
>
> We all might think it is a good idea but I for one am not going to 
> condemn a data publisher if it uses another scheme or has not put up a 
> document at some URI at the time of publication.  The publisher might 
> have very good reasons for it.

Not disputing any of that. My issue is that none of the above justifies 
conflating RDF and Linked Data.

>
> From experience, this is impractical anyway because you can loose 
> control of a URI space.  URIs live on.

Yes, but the URI/URL address space isn't as important as the actual 
*denotation* an *indirection* functionality delivered by generic Linked 
Data URIs. This is why (of late) I've been publishing notes about sample 
Turtle document [1] that demonstrate how Linked Data can be deployed in 
a manner that showcases the power of entity names over entity 
description document addresses, without burdening document publishers 
with any of the following:

1. domain ownership
2. dns server access with admin privileges
3. web server ownership
4. web server access with admin privileges
5. URI style choices for entity name/ description document address 
disambiguation.

>
> JSON-LD does not mention http: schemes and is not limited to them. 
> http: is highly desirable but not necessary.

But I am talking about the RDF side of things here; specifically about 
the *inaccurate claim* that any RDF document is a Linked Data document.

Links:

1. http://bit.ly/RJzd9S -- DIY-style Linked Data deployment via Turtle
2. http://bit.ly/O4LNKf -- DIY-style LInked Data deployment and 
Web-scale verifiable identity.

>
>     Andy
>
>
>


-- 

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