Archives of the TeradataForum

Message Posted: Wed, 24 Jan 2001 @ 17:24:08 GMT

	<Prev	Next>		<<First	<Prev	Next>	Last>>

When FDL was originally written relational theory, according to Date & Codd (and long before there was any kind of standards), prohibited duplicate rows. [Remember that, with the exception of multiset tables, it's just not only FDL that eliminates duplicate rows: for example an INSERT or INSERT/SELECT also eliminates duplicate rows.] At the time when the original Teradata design decisions were being made, all the database providers were constantly citing Date & Codd as if they had the sole definition of relational theory.

Since that time, various vendors have permitted duplicate rows because it does improve the performance of their product. Subsequently, duplicate rows have been included in the ANSI standard - largely under pressure of those vendors who have already incorporated them in their products.

Why doesn't FDL load duplicate rows into a multiset table? Depends on how cynical you want to be. On one side, FDL works fine the way it is and where's the benefit in changing it? On the other side, are there enough customers who want to use multiset tables to justify the development effort? Only the NCR decision makers can answer that and it appears that for the immediate future, duplicate rows will be eliminated.

	<Prev	Next>		<<First	<Prev	Next>	Last>>