A Comparison of Three Systems Modeling Methodologies - High-Performance Web Databases: Design, Development, and Deployment - page 206

used when multiple occurrences are associated with a single relationship
instance. This gives added information that is not indicated by other models,
but they can be simulated by ternary relationship in other models.
A qualifier and ordering constraint are used in OMT to further clarify an
association. A qualifier can be used to reduce the multiplicity of an associ-
ation. An ordering constraint which is shown at the many side of an asso-
ciation shows that the object requires a specific order to be a valid relation-
ship. An example is the relationship between a Screen object and a Window
object. There can be many windows on a screen but it is important to know
which is the topmost window. To indicate this requirement an ordered con-
straint is placed above the relationship line next to the Window object.
Another special type of relationship that can be used in OMT is aggrega-
tion. Neither the Entity Relationship or Information models have this type
of notation. This type of relationship shows the strength of a parts-whole
relationship. An example is a Lamp object that is an aggregate of a Base,
Body, Shade, and Bulb objects. Aggregation in a typical ER model is more
limited, because it requires two entity types and one relationship to form a
high-level entity.
To verify the relationship in ER and Information models a correlation
table can be used. This is actually the implementation of a relationship. The
table lists the instances involved in an association. By checking these
instances the validity of an association can be confirmed. In OMT the follow-
ing criteria are used for determining unnecessary or incorrect associations:
• associations between eliminated classes
• irrelevant or implementation associations
• actions
• derived associations
It is also recommended to break ternary relationships into binary
relationships.
The generalization structure and approach used by all three models is
very similar. Both the ER model and Object Modeling Technique allow the
use of an attribute to distinguish the type of generalization. In the ER model
this attribute is called a classifying attribute. In OMT it is called a discrim-
inator. An example of this concept is shown in the Pay_Method generaliza-
tion structures in Exhibit 3 and 7.
All these models consider disjointness constraints. If generalization is
disjoint then each entity can only be a member of one subclass. If it can be
a member of more than one subclass then the generalization has overlap.
Elmasri & Navathe's convention shows this constraint in by placing a “D” for
disjoint or “O” for overlap in the circle of the generalization symbol.
Exhibit 3 illustrates the disjoint structure of the Pay_Method generalization.