A comparative analysis of methodologies for database schema integration

Reviewer: Csaba Joseph Egyhazy

Schema integration, as defined by the authors, occurs in two contexts: (1) view integration (in database design), which produces a global conceptual description of a proposed database; and (2) database integration (in distributed database management), which produces the global schema of a collection of databases. To understand the complexity of schema integration, the authors begin by identifying some of the causes for schema diversity. This is followed by a number of comparisons of 12 established schema integration methodologies. The issue of the diversity of data models is resolved by adopting a uniform treatment of concepts, based primarily on the entity-relationship model. The conforming of schemas problem then amounts to resolving type, dependency, key, and behavioral conflicts. The resolution of the conflicts leads to schema transformations. This is an activity usually undertaken by designers, in close collaboration with users. However, as noted by the authors, most of the schema transformations are geared for a removal of redundancy, as opposed to simplification or logical optimization. One of the most disturbing revelations of the paper, which I found particularly noteworthy, was the absence of existing specialized languages or data structures for automating at least some of the four major schema integration activities common to all 12 methodologies identified in the paper. Additionally, only a few of these methodologies provide explicit tools or procedures to carry out the process of resolution beyond renaming, redundancy elimination, and generalization. The more difficult ones, such as integrity constraints and language and data structure incompatibilities, remain, for the most part, unresolved. And finally, as noted by the authors: “none [of these methodologies] provide an analysis or proof of the completeness of the schema transformation operations from the standpoint of being able to resolve any type of conflict that can arise.” This leads to the conclusion that none of the methodologies are based on any established mathematical theory and are merely engaged in defining a consensus schema by possibly changing some user views. This approach, first suggested over ten years ago, should be challenged by the slow, incoming influence of applied database logic among schema integration researchers.