The founding fathers of the relational data base, by using a combination of mathematics and black magic, came up with a set of rules defining the make up of a pure relational database. Nearly all of these rules have been bent or ignored by relational data base vendors and implementers, and for a good reason, a compromise between purity and performance is essential in the real world. A good example of such compromise is in the degree of normalization that is implemented in a data base. Normalization is measured on a scale of 1 to 5 "Normal Forms" - a kind of Richter Scale of database purity. The first normal form states that one value in a table should be just that - i.e. it is acceptable to have a well name column, with one value per line, but not to have a well names column , with a group of well names in a single field. The second normal form states that a row in a column must be uniquely identifiable while the third eliminates redundancy of non-key information. While these three requirements combine to make for robust database design and simple maintenance, they can have a negative effect on performance. Because of this they are frequently only partially applied.