Re: ISSUE-66: LinkedData™

On 08/05/2014 05:07 AM, Ruben Verborgh wrote:
> Hi all,
>
> Jumping in, as this is very relevant for the Linked Data Fragments
> spec [1]. In fact, this issue appearing after I drafted an
> introductory section called “What Linked Data is” might not be a
> coincidence. (And it's very good timing in any case.)
>
> Let me start out by saying I was totally oblivious of ”non-RDF Linked
> Data”.

That's like saying you were totally oblivious to the existence of the 
non-URI based World Wide Web.   There's a good reason why you were 
oblivious to their existence: they do not exist!

> I.e., I had always assumed that Linked Data is in RDF;
> probably because Tim's original principles explicitly mention this
> [2]. Then again, we all know the principles are quite vague: - RDF*
> and SPARQL are mentioned between parentheses. Did this mean "e.g.,
> RDF*, SPARQL", or "i.e., RDF*, SPARQL"? That's an important
> difference, and we'll likely never know. - Where is the asterisk
> after RDF ever resolved? Maybe I just missed the majority of the
> discussion; i.e., posts like [3] were written in 2009.
>
> That said, me being in the community for 4 years and never having
> heard about (or being selectively deaf towards) non-RDF Linked Data,
> means something at least. I'd dare to say that the majority of people
> do assume that Linked Data is just done with RDF. So to what extent
> is it then necessary to clarify this?

Unfortunately, it has become painfully evident that there are a few 
people who do not realize that Linked Data implies RDF (or who wish that 
it didn't).  For this reason, I think it is important to be clear about it.

>
> In that context, Dan Brickley sent a useful comment to me: “RDF is to
> Linked Data as HTML is to the classic Web, maybe”.
>
>>> i was specifically trying not to get that discussion going. just
>>> asking whether there should be some definition/clarification of
>>> the term, just to let readers know what it means in the context
>>> of the spec/community. if you define a broad term to mean a
>>> narrow thing, then this might be helpful to avoid possible
>>> confusion.
>
> What do you think about the current introduction to the triple
> pattern fragments spec [1]? Not knowing about this issue yet, I
> phrased it as:
>
> By publishing Linked Data [LINKED-DATA], we enable automated clients
> to consume information. In practice, this information is available as
> RDF triples […]
>
> So it leaves the question open whether non-RDF Linked Data exists; it
> just says that, in practice, it will be RDF. Good enough?

No.  It is important to *not* leave that question open.  That was the
whole point of the huge debates that occurred about this -- debates that 
(thankfully) finally ended with the official publication of the W3C 
Linked Data Glossary.

>
>> I think a definition could help.  I suggest copying the one from
>> the W3C Linked Data Glossary verbatim (and referencing that
>> document), rather than trying to craft a new one and risking
>> another long debate about what it should be.
>
> Sadly, I think that definition is quite complicated. Here it is at
> full length, copied from [4]:
>
> Linked Data
>
> A pattern for hyperlinking machine-readable data sets to each other
> using Semantic Web techniques, especially via the use of RDF and
> URIs. Enables distributed SPARQL queries of the data sets and a
> browsing or discovery approach to finding information (as compared to
> a search strategy). Linked Data is intended for access by both humans
> and machines. Linked Data uses the RDF family of standards for data
> interchange (e.g., RDF/XML, RDFa, Turtle) and query (SPARQL). If
> Linked Data is published on the public Web, it is generally called
> Linked Open Data. See also [Linked Data Principles].
>
> It forces you to understand: - Semantic Web - RDF - URIs - SPARQL to
> make sense out of it.
>
> And personally, I wonder to what extent SPARQL is part of Linked
> Data; and does that mean the query language, the protocol, or both?

I agree that that definition is not ideal.  And maybe you could come up 
with a simpler definition that would be acceptable.  But please be aware 
that the term "Linked Data" is very important to the semantic web 
community, and there are people with strong feelings about it, so 
crafting an alternate definition runs a risk of long debates.

>
> On the technical level, nothing prohibits us from making Linked Data
> Fragments broader than RDF. We'd have to be very careful, however,
> that the concept would still be sufficiently meaningful; that it
> doesn't become hollow by broadening it.
>
> For triple pattern fragments, by definition, we are limited to the
> RDF triple model. That does not mean that other kinds of fragments
> would have such a strong dependency; so other fragment types we
> define might be independent of RDF.

I don't see a fundamental problem with that if there is sufficient 
motivation for it.  Even though LDF was designed for Linked Data, I 
think it would be okay if it also happens to work with *other* data. 
But it is important to avoid implying that there is any such thing as 
"non-RDF Linked Data", because that would just cause confusion and 
reignite unnecessary debates.

David

>
> Best,
>
> Ruben
>
> [1] http://www.hydra-cg.com/spec/latest/linked-data-fragments/ [2]
> http://www.w3.org/DesignIssues/LinkedData.html [3]
> http://cloudofdata.com/2009/07/does-linked-data-need-rdf/ [4]
> http://www.w3.org/TR/ld-glossary/#linked-data
>
>

Received on Tuesday, 5 August 2014 18:03:08 UTC