Different Implementations of Software patterns.

• A Software Pattern is meant to be a generic solution for a recurring problem. Then why does implementation of a pattern vary?

• .Net/Java Event mechanism, MQ Series/MSMQ messaging, COM Bidirectional communication mechanism Event-Sink etc all are based on Observer pattern. What forces them to have different implementation?

•  Is it not possible to create a generic implementation of a pattern that is configurable and fits to all contexts?

• Can we consider patterns as complete solutions?

           No says

Pattern Oriented Software Architecture Volume 5: On Patterns and Pattern Languages

By: Frank Buschmann; Kevlin Henney; Douglas C. Schmidt

• A pattern differs from a specific solution to a problem by a) Recursive nature of the problem b) Roles of components involved in a pattern are strictly defined. But the organization of components to perform the role differs based on context.

•  The structural , behavioral and internal implementation variations of the pattern are influenced by
  o   Context in which it is applied.
  o Forces like a) Extensibility b) Testability c) Language agnostic nature of components d) Performance etc.
  o   Roles involved in patterns are universally applicable, but implementation of those roles by different components varies based on the real context.
  
  
•  The main purpose of observer pattern is to provide a mechanism for change propagation.

•  Important roles involved in observer pattern are a) Subject b) Observer.
  o   Subject   :- Notifies registered dependents.
  o  Observer: - Dependents on Subject State. Whenever state of subject changes , they should perform necessary actions.

• Variations in observer pattern

o   A single object might be an Observer in one relationship and Subject in other.

o   A single subject may receive events from a single subject or from multiple subjects.

o   A subject may be observed by single observer or multiple observers.

o   Subject and Observers may be components in a distributed system separated physically.

                  o   Subject and Observers may be components in an Enterprise Integration Context.         
                      Subject needs to be unaware of its Observers and
                      Observers needs to be unaware of Subjects.

•  Non functional requirements of the problem context like; performance etc might also control the behavior. Example: - State propagation mechanism might be push or pull. The data structures used might be influenced by the time complexity of operations that will be performed.