Atomicity and Normalization

A common aim of data modeling approaches is to produce schemas whose instantiations are always redundancy-free. This is especially useful when the implementation target is a relational database. This paper contrasts two very different approaches to attain a redundancy-free relational schema. The ObjectRole Modeling (ORM) approach emphasizes capturing semantics first in terms of atomic (elementary or existential) fact types, followed by synthesis of fact types into relation schemes. Normalization by decomposition instead focuses on “nonloss decomposition” to various, and progressively more refined, “normal forms”. Nonloss decomposition of a relation requires decomposition into smaller relations that, upon natural join, yield the exact original population. Nonloss decomposition of a table scheme (or relation variable) requires that the decomposition of all possible populations of the relation scheme is reversible in this way. In this paper we show that the dependency requirement for “all possible populations” is too restrictive for definitions of multivalued and join dependencies over relation schemes. By exploiting modeling heuristics underlying ORM, we offer better definitions of these data dependencies, and of “nonloss decomposition”, thus enabling these concepts to be addressed at a truly semantic level.