Thanks for the responses Helen!

> You can't really calculate it that way, because the indexes
> are compressed bitmaps: searches will "glide over" matching
> sequences. It's an issue when keys are larger but a 20-byte
> key wouldn't cause a significant difference.

Ann beat this into my head... and we changed our key creation routine
to place the most static elements first, to take advantage of the
previx compression

> If you are going to use GUIDs (and plenty do: it's a good
> solution when replication is involved) then use character
> set OCTETS so that you never need to bump into
> transliteration problems with keys.

A probably short sighted question here, but if this database will
_never_ be used to hold international chars, is there a benefit to
changing the database's overall default charset to OCTETS? - or am I
missing something here?


Thanks again... I sure am looking forward to moving forward to FB+IBO
and leaving BDE+ODBC+MSSQLServer!!!!
--Raymond