Caution — Relationship Work Ahead
In all of the discussion about detecting, repairing, and eliminating causes of data failures, the issue of relationship is mentioned rarely, if at all. If mentioned, it is always in the context of “relational integrity,” which has come to mean verifying that a “foreign key” value is traceable to an actual record in the “related” table. This job is often delegated to the DBMS.
This being the case, all that needs to be done (so goes the current thinking) is to document a relationship in the data model, from which the database schema is generated, and make sure integrity checking is turned on. This article shows why this approach is destined for failure and suggests some ways to uncover and eradicate this most subtle, most devious, and most costly of all data quality problems.