> >I cannot understand why anyone would
> > want to run a statement without a where clause on a million-row
> table
> > (well, possibly if there were some kind of grouping reducing the
> > returned set to a few rows)
>
> My query does only return a few records from the table, I'm using
> a 'distinct' clause.
>

so you want to scan 2.1 million records to pull back 5 distinct occurrences
(handfull)? and you're holding the index efficiency as responsible for poor
timing?
Seems to me that a bit of normalising of your data is on order no?
Alan