How did a 1970 paper about tables come to run the world's data?

A 1970 paper by an IBM mathematician named Edgar "Ted" Codd proposed a simple idea: keep data in plain tables, and let people ask for what they want instead of spelling out how to fetch it. That idea, the relational model, now sits under most of the databases in the world.

Before it, databases were navigational. Records pointed to other records, often by physical disk address, and programs followed those pointers like a treasure map. Codd's model said applications should search for data by its content instead of following links. Data lives in tables, one for each type of thing, such as customers or orders. A key is the smallest set of columns that tells the rows apart, and a row in one table can refer to a row in another by a matching value. To bring information together from several tables, you join them.

The surprising part is that IBM, Codd's own employer, resisted. It already sold IMS, a hierarchical database, and it refused at first to build a relational one in order to protect that income. Codd went to IBM's customers and showed them what the model could do; they pressured IBM, and rivals began building their own. By the mid-1980s computers were fast enough to run relational systems widely, and by the early 1990s they dominated large-scale data processing.