Should Variation Be Encoded Explicitly in Databases?

Variation occurs in databases in many different forms and contexts. For example, a single database schema evolves over time, data from different sources may be combined, and the various configurations of a software product line (SPL) may have different data needs. While approaches have been developed to deal with many such scenarios, particularly in the fields of database evolution and data integration, there is no solution that treats variation as a general and orthogonal concern in databases. This is a problem when various kinds of variation intersect, such as during the evolution of a SPL. Previously, we have proposed variational databases (VDB) as a general way to represent variation in both the structure and content of databases. Although the model underlying VDB is simple, encoding variation explicitly in databases introduces complexity akin to using preprocessing directives in software. In this paper, we explore the feasibility of VDB and its associated variational query language for encoding different kinds of database variability. We develop two use cases that illustrate how different kinds of variation can be encoded and integrated in VDB, and how the corresponding information needs can be expressed as variational queries. We then use these use cases to discuss the benefits and drawbacks of such a direct encoding of variation in data and queries.