Versioning Relational Data: a lot easier than you may think

Almost everyone who writes code is familiar with source control systems which track the changes to source code over time. Within relational databases, however, about the best way to track changes to data is some form of history table. It turns out that all changes to specified data can be tracked, maintained and viewed with minimal changes to the underlying model and few if any changes to existing apps (except, of course, those apps that are designed to work with "past" data or the history tables themselves).

Make verisoning automatic, transparent and flawless, with no compromising of data integrity -- including referencial integrity, the big bugaboo of most versioning attempts. Get rid of all history tables and the triggers, stored procedures and app code used to maintain them. In fact, current data and past data are viewed with the same query.

For those familiar with bi-temporal concepts, versioning is basically just implementing transaction time. However, from versioning to full bi-temporal functionality is a small step. For those interested in full bi-temporal capability (transaction time and effective time) I will have have handouts available showing how to get the rest of the way.