Mike,

>We'll never know until M$ tells us. It makes sense to me that intermediate results be indexed so that further processing within the SQL interpreter can find matching records.

I've got the following theory:
Since mutlitable query's are mutli pass processes (you've got to scan trough two ore more tables or indices) it make's sense to index table expressions (I'd rather not think off intermediate results since I don't think there are intermediate results in a singe SQL statement unless maybe using a subquery).

For in a single table action I cannot see any benefit to ad-hoc index expressions because the indexing process has at minimum to retrieve the actual record from the table to know it's contents, thus slowing down performance. Secondly where does rushmore want to use the index since the actual record is alread read for the indexing process.

Of course it's true we don't know the exact implementation of rushmore, but when we analyze the data we know of rushmore, it's highly unlikely this ad-hoc indexing is an part of rushmore. I don't think we should hide behind the fact that we don't know the exact implementation of rushmore, but make clear about the things we do know of rushmore. Since there still are many, many questions which are rushmore related I think that it would be time to do something about this.

Have any ideas ?

Walter,

>Thanks!