Apart from performance considerations, I prefer surrogate keys because they abstract the relationship of tables (columns used for RI) from the data itself. Then a separate unique index must be...

• This reply was modified 5 years, 1 month ago by  Regan Wick @sqlRecords.