Practical Guidelines for Supertype and Subtype Modeling - High-Performance Web Databases: Design, Development, and Deployment - page 321

Object Proliferation
The greatest problem that data modelers immediately encounter when
they conduct subtype modeling is object proliferation in the model dia-
grams and supporting documentation. A nontrivial diagram may contain
several subtype entities for one supertype. Each of the supertype's
attributes and outbound relationships may be replicated for each subtype
entity to constrain the cardinality or domain. When a supertype entity has
five attributes and is involved in three relationships, this means an addi-
tional 120 objects (i.e., subtypes, attributes, and relationships) in the data
model. The model may now contain 60 new attributes; however, only five
fields that correspond to the supertype's original five attributes may be
implemented. The knowledge that the data modeler gains through subtyp-
ing adds precision to the defined use of the data but does not necessarily
contribute any new data.
Although the data analyst may find these initial stages of subtyping to be
cumbersome, further investigation often uncovers additional subtype-spe-
cific data requirements that may be overlooked without subtype modeling.
This new information provides value to the project, but is only useful if ana-
lysts allow the number of entities, attributes, and relationships associated
with the model to grow rapidly.
Arbitrary Decomposition
Another problem with subtype modeling is the selection of appropriate
subtype perspectives. Subtyping is primarily a decomposing technique.
Successful decomposition requires a clear and unambiguous criterion:
Mixing the criteria within a single attempt at decomposition may lead to
results that do not isolate subtype knowledge in a single location in the
model but distribute it incorrectly across multiple points in the diagram. In
addition, the selected decomposition criterion may not be relevant to the
project. This results in the proliferation of subtype entities that contain no
attributes and cannot participate in relation- ships because they do not
require the constraint of any subtype involvements.
Conflicting Hierarchies
The decomposition associated with subtype modeling is hierarchical in
nature. Subtype relationships that are more network-oriented are difficult to
define when the data modeler uses techniques. Therefore, it may be difficult
to provide for all known data constraints with current subtyping techniques.
List Variability
Subtype modeling is difficult to implement with subtype schemas that
cannot be exhausted, either because the entire subtype list is unknown or
because the list may grow over time to include undefined types. In this