In a relational database the data is arranged in relations which are nothing but conventional tables with some set of rules governing the arrangement of data within the table and among the tables. Any record or any row in a table is called an entity. And any value (or cell) in the row is called is called attribute. Now the set of rules (12 rules) governing the relational database are called Codd's rules (proposed by Dr. E F Codd in 1970 (?) ). To put things simple, a) An entity must have a unique set of attributes, that is, every row in a table must be unique. b) The referential integrity between two relations must be preserved. That is, if a record in a table A depends on a record in table B, then the database must detect and cancel any operation in table B that disturbs this integrity. e)Domain integrity must be preserved. That is, if there is a restriction for an attribute set (column) that its value should be within a predefined range then this restriction must apply in every insert/modification. d) Any value which is not supplied or unknown should be termed as null. A null value is not equal to empty syring ("") or zero. A null is even not equal to a null. e) Access to the database should be easy, but the ability of the user to view or modify the data should depend on permissions. The part of the database for which the user has no kind of permission is invisible to the user. To know further, read discussions on Codd's rules .

There are 2 types of database: flatfile (everything is stored in one big table - like a telephone directory). Relational: 'Like' data is grouped together into seperate tables. Eg: A Library Database may have a table storing Customers Details a table storing Book Categories a table storing book details a table storing lending records The tables are related together using a Primary Key which is a unique identifier for each record. Eg: Books Table BookID (unique field) BookName text field A relational database minimises data errors, redundancy and duplication (each record should be stored only once (eg, when a book is lent to a customer, you would not want to have to enter the customer's details each time. Instead you store their details once in the Customers Table and then each time a book is lent to them, you look up the record required. Promotes data integrity (in other words the data stored is accurate) which means data storage costs are minimized.