On Thu, 2005-05-12 at 07:48, Job Miller wrote: > > a 20k row fetch via an index would be a bad idea.
Without knowing more about the data and schema, you can't really claim that. If the 20k rows are from a billion row table that is not partitioned, an index may be a good idea. It depends on how scattered the data is (the clustering_factor), how many index block reads it will take to find all rows, and how table blocks need to be read. Just use the optimizer and let the database do its job, it will save you work. :) Jared
