This implies no duplicate rows. Relational systems work on sets of rows distinguished by data value. It is impossible to distinguish duplicate rows, making it difficult to manipulate the duplicates independently. dB2K handles duplicate rows better than many systems. Still, try writing an SQL query that only deletes one of two duplicate rows! Some other relational systems even throw the extra row away as useless and meaningless. Save yourself trouble, design your database so that you don't permit duplicate rows in your tables.