Selection of reverse engineering methods for relational databases

The problem of choosing a method for the reverse engineering of relational database systems is not trivial. Methods have different input requirements, and each legacy system has its particular characteristics that restrict information availability. In this paper, we propose a classification framework based on the method's input requirements, namely: attribute semantics, attribute name consistency, data instances, applications source code, candidate keys, 3rd normal form (3NF), inclusion dependencies and human input. We present a short description of seven reverse engineering methods and analyse the applicability of each method to existing database systems, as well as the possibility of applying different methods to different parts of the system.