About Modeling Conventions
Using standard approaches—“conventions”—in the way we approach our work greatly improves communication among analysts, and it makes the analytical work easier. Because conventions establish a framework, entire categories of decisions do not have to be made. Until now, the data modeling field has not yet been mature enough to have established a complete set of standard practices. While much has been written about the syntax and grammar of data modeling, precious little has been written about the other elements that make up good modeling standards. In fact, three levels of convention apply to data modeling:
- Syntactic conventions are at the most basic level of defining models. These conventions dictate the symbols to be used. The symbols portray the things of significance to the enterprise (entities), and the relationships among them. Relationship symbols include those for cardinality (the “one” and “many” in a “one-to-many” relationship) and optionality (whether or not an occurrence of one entity must be related to an occurrence of the other entity). Many notation schemes also include symbols for describing how occurrences of entities are identified, even to the point of specifying the primary keys that would be used in relational database implementations of the models. Syntactic conventions may also include rules for the structure of phrases used to name relationships and entities. Syntactic conventions are the subject of most data modeling books.*
- Positional conventions define the way symbols are organized on a page. That is, they concern the relative positions of elements, and they address the overall organization of a drawing.
- Semantic conventions have to do with the grouping of entities according to their meaning. These are the primary subject of this book. These conventions pertain to the way we think about common business situations.
The three levels of conventions in data modeling are discussed more fully in Chapter Two. Here, it is important only to point out that it is through establishing conventions at all three levels that the full power of data modeling can be realized. It is the third level, the “conventions of thought.” that has been given the least attention, and which is the focus of this book.
Taking this pattern-based approach to data modeling has several advantages:
- It makes the task of building a new model easier and faster (and therefore cheaper, which after all, is what makes us popular with our bosses), since the modeler has only to modify and blend existing structures, rather than create new ones from scratch.
- It makes the models easier to read, since the same kinds of things will tend to take the same shape in all diagrams.
- This in turn makes it possible to highlight those things that are genuinely unique about a particular enterprise.
- It helps reduce or eliminate gross modeling errors, since the basic elements of a model are already defined.
- It will result in a system with fewer tables to maintain, simplifying it and making it more reliable, since relatively few entities can describe many specific aspects of a situation.