Are there arguments against ER models and their extensions


UML is more natural than ER

Opponent: UML is used everywhere. People cannot be wrong. UML has pushed away all other approaches.

Proponent: This might be true. A similar observation has been made in the beginning of the 80ies ... and the programming language PL/1. But now, PL/1 systems are considered to be bad legacy systems. You know why? Because of the deficiencies of PL/1. Since PL/1 had holes in semantics a large variety of compilers have been developed. Thus, the same program in PL/1 was differently translated in different environments. And had different behavior. Now, the CS community has learned the lessons. And may learn how to develop UML to a language which has a better fate.

Opponent: Sorry, but now you are becoming religious! And this reminds me the discussions of proponents of the relational model and the network model in the 70ies. Both side were becoming warriors against the other. And now you are going to a warrior.

Proponent: Yes, this might be an observation. But my intention is not to fight. It is rather to persuade you.

Opponent: Please give me one point why we should discuss our differences further!

Proponent: UML only uses binary relationship types. See also begin of discussion.

Opponent: You claim that UML is a language but not a modeling approach. What do you mean with this statement.

Proponent: Most natural language have a syntax based on an alphabet and a construction grammar. Programming languages as well. Words have a meaning in natural languages. This meaning can be used for the construction of the meaning of larger units such as sentences (if the language is constructive, e.g. most of the European languages). Computer scientists have learned the usefulness of construction of semantics in a similar fashion. Most programming languages are constructive (See for instance for Java: R.F. Stärk, J. Schmid, E. Börger, Java and the Java Virtual Machine, Springer 2001). One missing part is however pragmatics or pragmatism. Semiotics requires the consideration of the triangle: syntax, semantics and pragmatics. All natural languages constitute of syntax, semantics and pragmatics. Not so, most of the programming languages. Good programming languages have led to good methodological guides which allows to cope with pragmatics. Database modeling approaches do so as well. UML has a long way to go to reach this maturity. We claim that UML might learn this through HERM or through advanced ER models. UML has not a unique way of defining semantics. We miss methodologies in UML.

Opponent: Where I can read a good discussion on UML and a consistent comparison of UML, ER and OO models?

Proponent: Please go to the homepage of Klaus-Dieter Schewe and download his paper "UML - A Modern Dinosaur" presented at the European-Japanese Conference (EJC) 2000 on Information Modeling and Knowledge Bases (IOS Press).


`Other related information you find in page':More readings