The need for RDF in Linked Data

There seems to be some persistent misunderstanding about the role of RDF 
in Linked Data, as evidenced by comments like the following:

   "RDF is just one implementation of Linked Data"

If Linked Data is intended to support the goal of the Semantic Web, then 
unless the Semantic Web is re-architected with a new foundation, RDF is 
*essential* to Linked Data -- not optional, and not merely one potential 
choice among many.  the reason is that the Semantic Web critically 
relies on the use of *both* a standard universal identification 
convention (URIs) for its vocabulary, *and* a standard universal 
information model (RDF) for making statements.

To understand why a standard universal information model is important, 
one must think back to the central goal of the Semantic Web.  the goal 
is to enable computers to do more useful things for us: to enable them 
to find, share, combine and make meaningful use of web data.
http://www.scientificamerican.com/article.cfm?id=the-semantic-web
This means that a client application processing some web data should be 
able to follow links in that data to find more data that it can *also* 
meaningfully use.  the only way that can be achieved is by using a 
standard universal information model.  otherwise the client application 
will have no reliable way to properly interpret that new data.

For example, suppose the client application dereferences a URI and 
obtains a comma-separated-values (CSV) document.  unless the client 
application knew how to interpret that file, it would not be able to 
make meaningful use of that data.  it would be stuck at a dead end.  but 
if the document were expressed in a standard universal information 
model, then the client application would at least be able to understand 
what statements the document was making.  and if the client application 
did not already understand the vocabulary -- i.e. the meanings of the 
URIs -- then it could recursively, using Linked Data techniques, 
dereference the URIs to discover their meanings.

Why does RDF need to be the standard universal information model?  not 
because it is perfect, but because *some* standard universal information 
model is needed, and that is the one that was chosen, just as URIs were 
chosen to be the standard universal identification convention. 
furthermore, because RDF is syntax independent, a document does not have 
to *look* like RDF in order to be interpreted as RDF. for example, GRDDL 
allows arbitrary XML to be interpreted as RDF.  The enormous value of 
JSON-LD is that it provides a more web-developer-friendly syntax than 
ever before for a universal information model.

why couldn't other sufficiently powerful information models achieve the 
same Semantic Web goal just as well, and be used in addition to RDF? 
Because that would fragment the web.  instead of one web we would have 
many webs, each one its own walled garden, and that is not be Semantic 
Web goal.  without a shared information model, client applications would 
not be able to meaningfully combine the data from those walled gardens.

I do not expect anyone to take my word for this. All I ask is that you 
think about it.  Because if you do, the conclusion is unavoidable: if 
Linked Data is going to support the goal of the Semantic Web (without 
re-architecting it), then Linked Data MUST be based on RDF.

this obviously begs the question: *should* Linked Data support the goal 
of the Semantic Web?  that certainly was TimBL's intent when he coined 
the term and wrote his article about it:
http://www.w3.org/DesignIssues/LinkedData.html
(but I continue to be amazed at how differently different people seem to 
understand that article, so I imagine there would still be some who 
would disagree even with *that* point.)

*I* certainly think that Linked Data should support the goal of the 
Semantic Web.  and I think that JSON-LD -- *because* it will be such a 
web-developer-friendly RDF syntax (assuming a few small issues are 
resolved, so that it really *is* an RDF syntax) -- will be a big step 
forward.

If the term Linked Data is "hijacked" by a broader population to mean 
*any* sort of data that is linked -- not necessarily RDF -- then this 
will be a major loss to the Semantic Web community, because it is very 
hard to come up with simple ways to communicate the essence of the 
Semantic Web.  The Linked Data meme has been extremely helpful.  If the 
RDF component is lost, we will have lost the best meme we have ever had 
for explaining the Semantic Web.

David

Received on Monday, 17 June 2013 05:27:10 UTC