Database Reference
In-Depth Information
6.
Interaction increases Complexity
: ∃
P
, ∃
Q
, such that µ((( ) + µ(
Q
) < µ((( +
Q
). The
principle behind this property is that when two classes are combined, the interaction
between classes can increase the complexity metric value.
We will now consider each of these requirements on our cohesion metric, interchanging
the class notion for that of an activity. For this evaluation, we will assume that in general an
operations structure will contain a non-trivial number of operations which are not totally un-
related (i.e., there are at least two operations that share at least one information element).
Noncoarseness
With respect to the fi rst property, noncoarseness, suppose that there is an activity
t
defi ned on a certain operations structure. If
c
(
t
) equals zero, it is always possible to add two
related operations such that
c
(
t
) increases. If
c
(
t
) is unequal to zero, it is possible to take op-
erations away from it until its cohesion equals zero. Therefore, noncoarseness is guaranteed
on theoretical grounds. More practically, taking away or adding an operation will
mostly
affect both the relation cohesion and the information cohesion of that activity.
Nonuniqueness
If we consider the second property, nonuniqueness, it is immediately clear that each
activity with only one operation—regardless which one—has a cohesion of 0. So on theo-
retical grounds this property is satisfi ed. Practically, it will often be possible to defi ne two
activities with the same cohesion if an operations structure is suffi ciently large. For instance,
it will then be possible to fi nd two sets (activities) of two operations each, such that both
sets involve an equal number of information elements of which only one information ele-
ment is shared among the two operations it consists of. Both these activities will exactly
have the same cohesion.
Design Details are Important
An evaluation of the third property, design details are important, requires us to evaluate
the notion of functionality within the context of process design. We assume that two activities
are functionally the same when they share the same outputs, i.e., they contain operations
with equal outputs (which are
not
used as inputs by other operations within these activi-
ties). In practical operations structures it will often be the case that there are two operations
with the same output, but with different inputs, e.g., determining somebody's suitability
for a job by an interview or by means of a psychological test. The specifi c choice for one
of the alternatives to include in an activity will very likely be of infl uence on its cohesion,
satisfying the property in question.
Monotonicity
The fourth property of monotonicity is in general
not
satisfi ed by our notion of co-
not
hesion. There is a very good explanation for this: The explicit intention of the cohesion
metric is to decide whether it is wise to combine activities or not. If cohesion would have
been a monotonic property, combining activities
always results
in a higher cohesion. This
would have made the criterion worthless for our purpose. The original thought behind this
monotonicity property may be inspired by such simple complexity metrics as 'number of
not
satisfi ed by our notion of co-
