01/11/2011 | jiangws Without going into a dissertation on data modeling(itself a subject of many full-length treatises), tables should follow basic rules of normalization: Avoid duplicate data, avoid repeating groups in tables, and only put data in tables where the information is directly relevant.Remove Unnormalized relation ->(Remove repeating and multivalued atrributes)-> First normal form relation ->(Remove partially dependent attributes)-> Second normal form relation ->(Remove transitively dependent attributes)-> Thrid normal form relation ->(Apply additional normal forms) -> Fully normalized relation The need for Normalization: – Insert Anomaly – Delete Anomaly – Update Anormaly -> The mantra of the skilled database designer and data modeler is this: For each attribute, capture it once, store it once, and use that one copy everywhere. The process of normalization provides a certain degree of data protection by ensuring that repeating information is not duplicated to a great degree. However, this process also make the queries written to retrieve the data more complex and potentially harder to write since the addition of tables to a query requires more JOIN statements or WHERE clauses. So we use VIEW to simply the writing of SQL.