>> Would that statement not be equally slow if the column was plain ASCII?
>
> Slow, yes. Equally, no.
>
> In order to get to 2000th character in ASCII string, you simply point to

> start + 2000bytes. With UTF8, you'd have to read all the bytes and count

> the characters one by one until you reach 2000th.

I think that this processing overhead is minimal compared to all the other
work that the CPU must do to process such a statement, but I understand
your point now. Thanks!

Paul