Database Reference
In-Depth Information
The difference comes when we wish to add a road but do not know the name or
realize that for another road it is also important to store the road's old name as well
as its current name. The closed world solution is unable to handle either of these
cases without changing the database schema and constraints. We cannot store a road
unless it has a name, so if we wish to store a road where the name is unknown, then
we have to relax the name constraint. In the case of adding new data, we must physi-
cally alter the database structure to allow this new data to be held. In an open world,
neither challenge is problematic. We are able to enter the road with the unknown
name as the constraint just tells the system that the road has a name; we just do
not know what it is. Adding a new item of information is also straightforward: a
new predicate say,
has _ old _ name
is created and used to reference the road's
previous name. No change is required to the data structure; the structure was triples
before and remains triples afterward.
Of course, there are times when we want a closed world, especially if we are deal-
ing with data internal to a single organization. Here, an open world approach can be
more than a little frustrating. From an initial purist standpoint, there is increasing
recognition that there are times when it is useful to “switch off” open world reason-
ing and allow standard database constraint rules to take over. In the future, it is
therefore likely that tools will increasingly enable users to toggle between closed
and open world approaches. Initial mechanisms for this are appearing in tools like
Topbraid SPIN, discussed in Chapter 8, which harnesses SPARQL queries to test for
Linked Data validity.
The following are points to remember:
•
The open world assumption is a very different way of viewing information.
•
The open world assumption allows us to cope with the boundless extent of
the Web.
•
It differs from a closed world solution by requiring us only to specify mini-
mum characteristics of things, not everything we think we need to know.
This means we can vary the information we hold about specific feature
types, we can add more things about specific instances, and we can work
with incomplete data.
•
Sometimes, it is useful to enforce closed world methods, particularly when
we want to engage integrity checks on internal data.
11.4 THE SIMPLICITY AND COMPLEXITY OF THE SEMANTIC WEB
One of the great strengths of Linked Data is its data structure and general simplicity.
Once people have understood the idea behind triples, the concept of Linked Data
as a graph or network of interconnecting triples is very easy to understand. Those
coming to RDF from an XML (eXtensible Markup Language) background will have
to look past the tags to see the graph, but the basic concept is pretty easy once it has
been grasped. The great elegance with this solution is that this data structure is both
simple and universal: All Linked Data is expressed as triples of the form
subject
predicate object
. Subjects may represent classes of things or individuals, predicates
(also known as properties or relationships) establish the link between subject and
