Information Technology Reference
In-Depth Information
be based on the number of objects and the nature of objects. A construct
has a better minimality score than another one if it uses fewer objects to
express the same fact type. We also consider the structural and cognitive
complexity of the objects. For example, to represent a definite fact type, an
elementary attribute is clearly less complex than a compound attribute or
an entity type.
- Expressiveness: Like simplicity, expressiveness has a high impact on under-
standability in the ISO standard [12]. An early definition of expressiveness
was suggested by Batini et al. in [16], where the authors defined it as
the
richness of a schema
. In this paper, we follow the definition given in [15]
inspired from the previous early definition, and more specifically the sub-
concepts of
concept expressiveness
, that measures whether
the concept
[the
constructs]
of the schema are expressive enough to capture the main aspects
of the reality.
We relate the expressiveness to the fact that a type of facts
of the application domain is represented by a construct that naturally and
clearly refers to its nature. For instance, the fact that
domestic appliances
form a variety of
products
should be represented by a subtype/supertype
relation, provided the data model includes this type of constructs.
- Evolvability: We define evolvability as the ability of a construct to support
possible changes in the application domain and to trigger as little impact
as possible on the system artefacts that include (e.g., the schemas) and use
(e.g., programs and HCI) this construct. While previous quality requirements
address the understandability of the conceptual schema, this definition is
adapted from the
changeability
definition of the ISO Quality standard [12].
Being able to adapt the schemas following new application domain changes
is important, especially in the context of the MDE methods.
These three qualities synthesize some of the most important requirements for a
database schema. In addition, they can be formally defined through our frame-
work, in which they form the
D
component of a context. We will apply them
to evaluate and illustrate the taxonomy of section 5. Besides, these qualities are
not independent: a construct that increases the simplicity of a schema may lower
its expressiveness. They can be considered separately or combined in order to
reach a trade-off.
5 Defect Taxonomy
Teaching, modeling experience and schema analysis eventually allow to come up
with a set of good modeling practices. The latter are the structures an experi-
enced designer would most probably use to represent specific fact types of the
application domain. We observe that many designers make other, sometimes un-
fortunate, choices, so that design flaws may often be found in database schemas.
In addition, the fact that a definite construct should be used is obviously context-
dependent. Besides, a sound construct may not always meet all the requirements
stated for some databases.
