>If the table is wide, but some queries
>use only the primary key value, a scan of the nonclustered index will be
>faster than a scan of the clustered index.

Thanks. That's good info.