>Sounds to me as if you just volunteered to prove your guess <g>

It proves me wrong, but...

On a table of roughly 50K records, indexed once on machine and once on a local sequence, randomly seeking both tags brought a ratio of roughly 0.19 for Upper and 0.21 milliseconds for simple with local sequence. The times have to include the fact that my machine is way behind todays technology (100MHz AMD K5, 32M) and that each loop called a function to get a random name. I began with a table of zip codes (city field only) and then began adding records, but the numbers didn't change.

So it comes down to sort of equal - so it's just that the programmer has to type "Upper(" and ")" every once in a while, just to gain some minimal speed margin.

I've rechecked the help, but couldn't find any conclusive evidence on the old question - are other collating sequences used in Rushmore?