Database Reference
In-Depth Information
also have a number of
attributes
: Spatial, Temporal, Quality, Location, Metadata,
Thematic. A feature is not defined in terms of a single geometry, but rather as a
conceptually meaningful object within a particular information or application com-
munity, one or more of the feature's properties may be geometric” (OGC, 2011).
The OGC feature is a basic component of OGC standards and is an abstract rep-
resentation of some
aspects of
a geographic feature in the real world. The quotation
contains a number of key aspects about a feature. First, it indicates that a feature is a
data structure that is restricted to a single information or application community—a
closed world with a strict and limited scope. Second, the feature wraps a number of
attributes as a single tightly bound package. It defines
all
the attributes that the fea-
ture may have; some can be optional, but it cannot have additional attributes that are
not described by the feature's definition. Again, this is closed world. An OGC feature
represents a data abstraction of a real-world object rather than the object itself. Each
OGC feature has a unique identifier, the Feature Identifier or FID, and this identifier
is associated with the OGC feature, not the real-world object it describes. This is in
stark contrast to a Semantic Web identifier (Uniform Resource Identifier, URI) that
explicitly represents the real-world object, is not bound to a particular community,
and identifies an object that may be incomplete in data terms. The difference in rep-
resentation between abstract and real features is important and often causes much
confusion, as it can be quite a subtle difference. Thinking about something as an
abstract representation is good when there has to be agreement between parties about
what is being referred and works well with internalized models; identity is applied
to this abstract data. One problem with this view is that ultimately the abstract has
to be grounded in the real, so in certain circumstances it can be seen as an unnec-
essary complication. A viewpoint of the item of interest as the real-world object is
better suited to information exchange and to an open Web. Here, identity applies to
the real thing.
1
It is not that the Semantic Web does not have abstract views; it does,
and these are represented in ontologies. The difference is more about how they are
presented. In the traditional database or object-oriented (OO) view of the world, the
abstract is bound exactly to the data. The abstract model represents the informa-
tion that a particular application requires, and this dictates the data structure that is
needed to hold that information. In the Semantic Web, the abstract model represents
a minimum classification for something and is held in an ontology, separate from the
data. Therefore, it is possible to represent more about a real-world object than may
be contained in any one abstract model. For example, an ontology may state that the
minimum requirement for a school is that it has a building with the purpose of pro-
viding education. In the Semantic Web, this does not prevent us from holding addi-
tional data about any specific school, perhaps saying it also has a car park and sports
fields. The abstract model defines minimum membership for the class “School”; the
real world is more complex, as is the Semantic Web data. In contrast, the traditional
abstract model provides a complete application definition, so if the application needs
to know that schools can have car parks, it had better say so.
In a pre-Web world or when an organization is interested only in an internal solu-
tion (or one that is exposed to parties that all share the same model), the closed
world assumption is perfectly valid. It is also efficient because it is able to enforce
strict compliance; data that does not conform to the model can be rejected and the
