Internet Archaeol 15. Nick Ryan. Data models

2. Data models

All DBMS are designed to support a particular data model, an abstract structure that provides a framework for organising data. Most of the systems encountered today are based on the Relational data model. In part, this is because the development of database use in archaeology over the last twenty years has followed the rise of this type of system to the position of dominance that it has now enjoyed for more than a decade. Previously, many large commercial and administrative organisations had used systems based on hierarchical and network data models, but these lacked the flexibility of their relational successors and usually required a programmer to write special programs to perform even the most simple input, update or retrieval. Fortunately for the archaeologist, these systems are rarely encountered today.

A key element in the success of relational systems was the early adoption of interactive, or ad hoc, query languages and, later, the ongoing development of SQL as an ISO standard query language which is now supported by the majority of modern DBMS. The wholesale adoption of this standard means that both user skills and data access software can be readily transferred from one DBMS to another. Minor differences arise as manufacturers seek to gain competitive advantage by adding new features to their particular implementations of the language, but this has also had the effect of driving the development of the standard to include capabilities far beyond those provided in early relational systems.

During the period when relational DBMS came to the fore, the Object-Oriented (OO) paradigm was gaining importance in the design of computer programs and in the development of programming languages. This approach now dominates these areas and has also had a significant impact on data management and database design. Through much of the 1990s, groups within the database research community debated whether the future of databases lay in the development of 'pure' object-oriented systems or in the extension of existing relational systems. Perhaps it is still too early to judge the outcome but, for the moment, the latter group appears to have won the argument. Whilst there are a few pure object-oriented database systems, many relational systems have been extended to handle a much wider range of data types and to support some of the elements of object-oriented models.

These developments, many of which have been formalised in the latest SQL:1999 standard (ISO/IEC 9075:1999; Eisenberg and Melton 1999; Melton et al. 2001), have resulted in several modern DBMS combining beneficial features of both relational and object-oriented approaches. Originally referred to as extended or extensible relational systems, the term Object-Relational (OR) has now been widely adopted to describe them.

The next three sections outline the main characteristics of these data models and illustrate how these characteristics impinge upon the process of designing and implementing a database.