Are there arguments against ER models and their extensions


Binary relationship types cope with anything what is necessary

Opponent: N-ary relationship types are not necessary in modeling.

Proponent: Consider the natural language. You can express anything what has to be expressed in the SPO form. The subject and the object are in this case the two participating entity types. The predicate or the verb expresses the relationship between the subject(s) and the object(s). If you need to express more complex associations you have to transfer these into SPO sentences. That is an experience every stranger has while using non-native languages such as Germans while speaking in American English. However see for instance a sentence from narratives. Another observation is the valency of verbs in natural languages. Let's look into a good language dictionary. You won't be able to find so much verbs which have a valency of 1 , i.e., having only one object. Of course you may use ellipse by reducing the valency. But during modeling we are interested in safe concepts.

Opponent: Sorry but oral communication in native languages is different. People are using a different style while communicating orally.

Proponent: Let us have a look into an oral sentence, for instance, taken from one of the great movie plots: In the great screenplay WITNESS by EARL W. WALLACE, WILLIAM KELLEY from a story by WILLIAM KELLEY your find find the following title sequence describing the situation:

The faces of several young children are presented in CLOSEUP, as they walk TOWARD US across a ploughed field.

Opponent: Sorry but this is not a good example of oral communication.

Proponent: Then let look further into the screenplay:

ELAINE : I know exactly what you meant! Listen, maybe I'm not a world-class housemaid, but maybe I don't have time to polish the goddamn china and keep after the kids! (and) It's none of your business, but I don't happen to have a man around here full time. So I sell cosmetics in a goddamn drugstore and sometimes I can even pay the rent on time! So maybe I'm not Mary Poppins, but maybe I don't need to have it jammed down my throat like this.

You may express this by SPO sentences. But then the narrative is much longer and much more difficult to understand. And nobody can capture what you said.

Opponent: Sorry, but as you stated we are able to represent these sentences via SPO sentences. We are interested in modeling in computational simplicity. and binary types are much simpler to compute and to maintain!

Proponent: This is true for structures. However, the association among the types remain. You have a very simple structure and you MUST pay the price via integrity constraints. In this case you have two options:

A) You introduce an additional type representing the association. This additional type is related to any of the components (let's say n) via n (!) binary relationship types. Additionally you must introduce n (!) mandatory cardinality types which have to be maintained via any insert/delete/update operation. This is far too much. And the efficiency of the binary types strikes back with superficial complexity of integrity maintenance. Or you switch off maintenance and you pay the price with data inconsistency (leading you back to the sixties, i.e. to the time before the advent of the databases. I have made the experience with students that developing and checking the corresponding trigger system becomes a true nightmare.

B) You introduce a surrogate attribute or an object identifier. This object identifier allows identification of objects in the the database if (and unfortunately only if) identifiers are value representable. Or you will loos objects in the database and you will never be able to distinguish objects in the database (then please count how many you will have).

C) You try to normalize as much as only possible. But there are still associations which are not binary. Therefore, go to B) or to A).


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