>>>As I understand Rushmore, it will be sucking the
>>>.CDXs across the WAN in both alternatives to determine which actual
>>>records to get.
>
>>The whole .cxd or idx is not transfered, only the nodes of the b+ tree that are >being traversed are being transfered.
>
>AHA! I *thought* Rushmore was quite clever. So why the delay opening the
>.DBFs and .CDXs?

Loading buffers with prefetch data so that after initial load, operations are faster.

In regard to your idea of removing the deleted tag, this will hurt performance if deleted is set to "on".

Open the tables at startup and keep them open until shutdowm.

>
>(To recap, the initial WAN test was FPW2.6 Windows, locally loaded
>.EXE; open several .DBFs with .CDXs, then do a full screen browse on the
>current file, i.e. BROWSE FOR with maybe a
>relation or two set to code->description fields.
>Time between each file open was around 45 seconds or more. Ages for the
>browse records to appear. Loading a remote .DBF into Lotus 123 was
>a *lot* quicker).