Microsoft Word OpenIssuesInOO. doc

2.2 

Focus on analysis and design 

In the late eighties, a number of object-oriented methodologies were proposed, reflecting 

that  modeling and design was gaining importance. We consider this development positive, 

since it encourages programmers to think about their designs before implementing tricky 

inheritance relations. In analysis and design it is obvious that subclassing should be used for 

representing specialisation of concepts instead of incidental inheritance. 

The methodologies do, however, seem to introduce some new problems to object-

orientation: 

• 

Their importance are often overestimated. A lot of effort seems to be used in drawing 

boxes and bubbles with little semantic meaning and often no code is ever produced.  

• 

New analysis and design languages are proposed. These languages are mainly graphical 

notations for expressing object-oriented language constructs like class, subclass, virtual 

procedure, etc. These new languages are often supported by CASE tools that support 

construction of diagrams and code generation. As discussed in section 2.3 this has 

reintroduced the traditional problems with CASE gap and reverse engineering.   

The real benefit of object-orientation compared to traditional structured analysis and design 

(SA/SD) is that object-orientation provides a unifying perspective on analysis, design and 

implementation. This unification also includes data modeling in databases, since object-

oriented databases provides much richer modeling capabilities than relational databases. 

Traditional SA/SD has never been used in the Simula/BETA community and probably 

the same can be said about most Smalltalk programmers. It has always been considered 

something that was needed when programming in COBOL, since there was a need for 

something at a more abstract level.  In contrast, Simula/BETA have been used for design 

since they were designed to be able to represent real world phenomena and concepts.  

It has always been difficult to teach object-orientation to people working with traditional 

SA/SD and Cobol, since they have insisted on having more than 'just a programming lan-

guage' for design and they did not accept object-orientation as useful in design. It helped 

when Yourdon together with Coad published his book on OOA [CY89]. This immediately 

made it much easier to advocate object-orientation. 

The problem now seems that there is too much focus on analysis and design. In addition 

people that used to be engaged in SA/SD methods are introducing problems in object-

orientation that did not use to be there. In SA/SD, the programming language has not been 

considered important partly because it is usually at a much lower level than the languages 

used for analysis and design. When converting to object-orientation, many methodology 

---

people continue to ignore the importance of the programming language. It is, however, the 

power of the programming languages that have made object-orientation so successful.