Purpose

Describe a successful phased approach to Unicode support mplemented for the Teradata very large database engine.

Description

In 1984, Teradata produced the first commercial highly parallel database engine. Since that time, many features have been added, including increasing internationalization support. When the decision was made to support Unicode, the database engine code was numbered in the millions of lines, much of it assuming particular representations for character data. It was clear that this was no easy task. In order to avoid a resource and risk bottleneck that would have been infeasible for a single release, a phased approach was employed. Each phase could be justified on its own merit, and acted as the groundwork for the phase to follow. Phase 1 of the Unicode implementation involved the construction of a Unicode datatype internal to the database engine. The byproduct and justification was the ability to interchange Japanese data between heterogenious clients with differing encodings for Japanese. Phase 2 constructed a method for defining external single byte characters, allowing appropriate character support for almost every nation in the world. Phase 3 externalizes the Unicode support, banishing the need for translation of incoming user data. Phase 4 regularizes Unicode support throughout the database engine, allowing fully internationalized object names. While the precise divisions between the phases may differ when planning Unicode support for other large existing applications, the Teradata experience provides a model for such efforts.

Conclusion

A phased approach to Unicode support within a large existing application offers an attractive alternative to a potentially overwhelming attempt to offer full support within the span of a single release.