On 06/11/2013 12:18 PM, Kingsley Idehen wrote:
On 6/11/13 11:56 AM, David Booth wrote:
On 06/11/2013 10:59 AM, Kingsley Idehen wrote:
[ . . . ] many RDF advocates
want to conflate Linked Data and RDF. This is technically wrong, and
marketing wise -- an utter disaster.
I have not heard RDF advocates conflating Linked Data and RDF, but
maybe you talk to different RDF advocates than me.
AFAICT, the vast majority of RDF advocates know that Linked Data is
RDF in which URIs are deferenceable to more RDF, but RDF is not
necessarily Linked Data, because RDF itself does not require URIs to
be dereferenceable.
David
RDF isn't the defining characteristic when speaking about Linked Data
outside the RDF community.
But RDF *is* one of Linked Data's defining characteristics, regardless
of whether people outside the RDF community understand that. (And it
seems to me that if they don't understand that, then we should help them
to understand that, rather than perpetuating their misunderstanding.)
It is much more palatable outside of the RDF
community to loosely couple Linked Data (the concept) and RDF (a
framework) which enables the construction of powerful Linked Data that's
endowed with *explicit* human and machine-comprehensible entity
relationships semantics.
We could define a new concept that decouples RDF from Linked Data. But
to avoid conflating Linked Data with this new notion, let's call this
new notion "Linked Stuff":
DEFINITION: **Linked Stuff** is data that uses URIs to
identify things, and those URIs are dereferenceable to
more Linked Stuff that describes those things.
There may be some good marketing benefit in adopting this simple notion
of Linked Stuff, since it avoids any mention of RDF, but there would be
a very important technical loss as well, in comparison with Linked Data.
To understand that loss, suppose a client application has some Linked
Stuff and wants to learn more about the thing identified by one of the
URIs in it. The client application automatically dereferences that URI
and receives more Linked Stuff. But can the client application
understand or do anything with the document that it receives? Not
necessarily, because that document may be in some random
tab-separated-values form, that the client application has no idea how
to interpret.
Compare this with Linked Data. In the Linked Data case, the client
application receives RDF. The document does not necessarily *look* like
RDF (to an untrained eye), because RDF is syntax independent, and there
are many different serializations of RDF. But as long as some kind of
standard RDF serialization was used -- even XML-based serializations
that use GRDDL[1] -- the client application is able to interpret and
make use of that document.
This is the goal of the Semantic Web: to enable machines to usefully and
(semi-)automatically, find, share, combine and process web data.
Because Linked Data is RDF, Linked Data supports that goal in a very
important way that Linked Stuff does not.
David
1. GRDDL: http://www.w3.org/TR/grddl-primer/
Why? Because you don't build friction with folks that are already
familiar with similar concepts albeit described using different
terminology.
The key is to build bridges rather than impede their construction by
enforcing world views in the most inflexible way.
If someone indicates to you that the letters R-D-F don't work for them,
for whatever reason, what's wrong with triangulation to the same
destination when it's the fundamental concept that matters, not the
labels that we slap on them at specific times in our innovation continuum?
RDF and the Entity Relationship model [1] outlined by Peter Chen in his
1976 dissertation are linked, conceptually and technically. That
association is very powerful and extremely useful in situations where
your audience suffers from R-D-F reflux.
RDF is useful, but it (like all innovations) has genealogy. That
genealogy is just as important as the innovations it adds to the continuum.
Links:
1. http://bit.ly/YTdz3N -- The Entity-Relationship Model -- Toward a
Unified View of Data (note: page 34) .
