While the implementation of the Data Manipulation Language (SELECT, INSERT, UPDATE, DELETE) has evolved into a fairly uniform implementation by the vendors, the implementation of data types has been in more disarray. Data-type implementation normally evolves as follows:

> An RDBMS vendor adds a new data type that is of value to a group of users.

> Other RDBMS vendors add the same or a similar data type with their own variation to differentiate their product.

> Several years pass and the data type grows in popularity with most vendors implementing a similar version.

> The standards committee gets involved to try to standardize the data type and eliminate variations.

> The RDBMS vendors slowly support the new standardized data type, but they must continue to support the users who have implemented their previous variations.

> Over several major releases from the vendor, users will migrate to the new standard data type and the previously implemented variation can be slowly phased out.

A good example of this process is DB2′s implementation of DATE, TIME, and TIMESTAMP.