Swen,
It really depends on the size of your table. If you have tens of millions of records then maybe SQL Server is the way to go. Otherwise, VFP can handle it quite nicely.

VFP doesn't actually bring down the whole index. Yair Alan Griver wrote something in FPA last year that explained it very well. VFP exaamines the index header to see if there are any it can use. It then downloads those indexes. This is usually quite fast.

If you are talking about using SQL Server then I would use SQL Pass-thru instead of ADO. The same filtered result will be returned but you will not have the added overhead of converting that to a VFP cursor (ADO recordsets can not be displyed in VFP grids directly).

HTH.

>I have an application with a central DB on an NT file server. The client machines display subsets of the tables in grids. For example, a user may want to see that's on-hand for an inventory item. Performance is so-so.
>
>My question is "How do I best select the subset of data?"
>
>It seems that since the application is running on the client, even doing a "select... where partno = 'abc'" would still have to bring the whole index over the LAN. It seems like there would be an advantage to using SQLServer as the backend and using an ADO Recordset so that the LAN traffic would only contain the actual records I want, since the selection is done on the server and not the client.
>
>Am I missing something here? Does anyone have any data comparing the 2 approaches?
>
>TIA
>Swen Nielsen
>garlo@centurytel.net