snip

>Helen has always said that the latter is the true state, i.e. a full scan

is

>performed before an order by clause starts filtering the return set but in

snip

> SELECT FIRST m doesn't make sense *unless* it's from an ordered
> set. The engine can't know which are the first m records of the full
> set until it has the full set and then has sorted it. If joins are
> involved, then the joins will happen first, since the content of set
> won't be known until the join criteria have been satisfied.
>

Helen - I assume an order by clause is used - I state as much
Alan