>In VFP it was not recommended because the index header might possibly rarely get corrupted. I never saw that happen in real life. But having the index code elsewhere eliminated that possibility.

That is correct. I didn't remember that one. Thanks for the info

>VFP adds new index keys and deletes existing keys during Create, Update and Delete. So there is some bloat. A reindex or rebuild both resulted in exactly the same size index file. I don't have SQL Server handy, but you could test that.
>
>Most importantly - creating the indexes via code (or Stonefield with VFP), lets you change the indexes automatically.

Yes, that has always been my approach.

Thanks