Modeling Business Rules - High-Performance Web Databases: Design, Development, and Deployment - page 142

enabling the database designer to work from standard rule components.
This simplifies the task of translation significantly.
EXTENDING DATA (AND OBJECT) MODELS
A successful approach to modeling business rules requires careful atten-
tion to several crucial objectives, as follows:
• Rules depend on specification of underlying data types; these data
types must be defined comprehensively and uniquely.
• To achieve maximum consistency and adaptability, rules themselves
should be specified in discrete and nonredundant fashion.
• Users generally are not concerned with how business rules are imple-
mented; therefore nonprocedural specification of rules is preferred.
• Communication of business intent usually is served best by pictures.
The framework that satisfies all these objectives most fully is provided
by data models. Object models also may satisfy these objectives. This is
discussed briefly in the following sections. Ross Method therefore uses
data models as a given in specifying rules.
This preference is grounded in fundamental beliefs about the nature of
production-level business operations. The computing problem this envi-
ronment represents is the extreme opposite of single-user systems (even
highly complex ones). The defining characteristic of the operational level
of a business is extensive, multipurpose concurrency. It includes signifi-
cant ad hoc access to persistent data by query language. The processing
requirements of different concurrent users often are significantly different
— and not infrequently, at odds with one another.
In environments of this type, ensuring consistency on the process side
(by embedding support for business rules in all the processes) ultimately
is futile. The only feasible solution is viewing specification of rules as an
extension of the database problem, and using data models (or possibly
object models) as the specification vehicle to achieve it.
Ross Method does not address data (or object) modeling per se. Rather,
it assumes a robust, state-of-the-art technique already is familiar. Such
technique must include complete, unencumbered support for type hierar-
chies. Any data modeling technique that satisfies these criteria may be
used. The rule-modeling approach of Ross Method therefore is intended as
a self-contained set of extensions to general modeling practices.
This includes many object-oriented (OO) approaches, assuming the
following: the given approach permits properties of objects to be viewed
publicly. (Certain OO approaches might be described as featuring hard-
core encapsulation. These permit no access to the properties of an object
except through messages to its operations. How declarative rules might
be specified under these approaches is unclear.)