> Now imagine this database is used by a Microsoft division, so that
> there are 3 million customers. That means the index will have three
> entries with million record pointers for each - and the corresponding
> number of database pages. Whenever IB has to add another record it
> has to insert a new pointer into that collection of a million record
> pointers - and it does not do this efficiently.
>

This means, that performance-problems will only happen on insert, update and
delete and not for a select ?

> However my main point is that you would probably not notice the
> performance impact if this happened on a database with only 10
> thousand customers. If it is not noticeable, then dont worry about
> it. If it is noticeable then...
>
> 1. Define your own index with a higher selectivity,

What means high or low selectivity of a index ?


Guido.