Picture a company's records in 1969. Data lives in files and in clever linked structures. A program that asks "who works in Sales?" does not just ask. It follows a chain of pointers from a department record, to its first employee, then to the next, and so on. If anyone reorganizes the storage, the program breaks.
In 1970, Edgar F. Codd of IBM published a short paper that changed that. His proposal: store data as simple tables, called relations, and let users ask for data by what it is, not by where it is stored. This is the idea behind SQL and every relational database since.
Describe data as relations (tables of rows with named columns), connect them by matching values, and give users a high-level way to ask questions. That separates what data means from how it is stored, a property Codd calls data independence.
The problem: programs welded to storage
Codd lists three ways that programs of his day depended on how data was arranged:
- Ordering dependence. A program assumes records are stored in a certain order. If the order changes, it gives wrong answers.
- Indexing dependence. A program uses a particular index to be fast. Remove the index and it breaks.
- Access path dependence. A program walks a tree or network of links. If the structure changes, the program must be rewritten.
Each of these made it expensive to improve a database. Every storage change meant touching every program that used it.
