Database is the strangest of computer applications. It is vital - every computer manages some sort of database, it is difficult - real world database quickly exceeds any reasonable complexity - and it is lucrative - more programmers are occupied with database work than any other.

Strangely, though, theories of how we should organize databases have been very thin on the ground. For such an important subject it has been relatively neglected.

The one exception is perhaps the work of E.F. Codd. Immortalized to a generation or two of database creators, he is revered as the originator of the commandment-like “Codd’s Rules”.

Edgar Frank Codd (1923-2003)

Codd was a mathematician and this is an important fact that explains much about his work.

He was born in the UK and studied math at Oxford before moving to the States in 1949 to do some work at the University of Tennessee. Before very long he was seduced by the new subject of computing and was working for IBM as a mathematician and programmer.

The first machine that he worked on was the SSEC - one of the early valve machines. He moved on to help with the logic design of the IBM 701, one of the first modern computers, and then the 702. At this point (1953) he took a four-year leave and went to study for his PhD in Canada, at the University of Michigan.

Codd’s work at Michigan was on a relatively new area, Cellular Automata. A cellular automaton is a collection of relatively simple computing units, all of which run the same program and interact with their neighbors on a rectangular grid. This was highly theoretical stuff at the time but it was connected with areas of AI and Artificial Life that are now very important. His thesis was on self reproduction of cellular automata.

Later the work was published as a book, “Cellular Automata”, in 1968. Many theoretical computer scientists know this book but often never suspect that the Codd who wrote it is the same one who did the eminently practical database work!

Database relationships

However Codd’s main work began in 1969 when he became interested in database software.

Being a mathematician he saw database in a very different way to the average programmer. A programmer tends to see the physical way that data is coded and stored. He sees records composed of fields of different types and the whole lot is stored as a file with some type of structure. The operations that are performed on a database are seen as a general programming problem. To work with a database you write a program that reads the file and works with it in any way that the user/customer considers necessary. If you need to access two records from different parts of the file at that same time it is up to the programmer to get inside and dig around the file in anyway they like.

This is a recipe for complexity to get out of hand! You can well imagine that the database file could be come altered in ways that are incorrect - even to the extent of it not being a valid database file any longer!

What Codd did was to see that the simplest example of a database was nothing more than a table of values. The rows of the table are records and the columns fields and the table itself can be implemented as a file. This change in viewpoint is, however, more than just a relabeling of the parts. A table, or set of multiple values in mathematical terms, is known as a “relationship”.

For example, suppose I ask you to tabulate the relationship x<y for the integers 1 to 5. What you end up with is a table of pairs - or in general “tuples”:

  x  y
  1  2
  1  3
  1  4
  1  5
  2  3
  2  4
  2  5
  3  4
and so on…

Any relationship, not just the nice neat regular ones like =, <>, < etc., can be represented by a set of tuples and this is exactly the way that Codd saw a database table and hence the name of the new approach “relational database”.

<ASIN:0201612941>

<ASIN:0321197844>

<ASIN:1425122906>