The goal of the optimizer is to reduce the number of pages read because
page reads are the most expensive database operation. An index on
a field that has two values evenly distributed is likely to reference every
page for each value, so the optimizer would probably ignore it. But the
threshold at which the optimizer discards an index changes. In the early
days, the people working on the optimizer assumed that database designers
knew what they were doing, and would use indexes even if they seemed
unpromising. Over time, that assumption has been scaled back, and is
likely to change with different versions of Firebird.

What a polite way of say DB developer's are getting a bit dumberer :)

-randall


[Non-text portions of this message have been removed]