One of the key differences beteen SQL type databases and "desktop" (paradox, XBase, etc.) is that there is no inherent order. Order is only established when you query.
You can only go forward and backward within the results of your query.
Picture a database with five million customers in it. You would not want to have that monster on your local machine just so that you can scroll one by one through it.
What is required here is a new way to see the problem. The data base is very large so we grab sections of it. We can move easily using GUI tools through that section but not the WHOLE database. To solve your immediate problem you might simulate the forward-backward behavior by returning a query that is wider than the target record by specifying a range. (User asks for Smith--the query returns ALL S's or All R's, S's and T's) Then the user can scroll the query results within this range. When user scrolls near the top or bottom of the range then a new query is generated and displayed.
Just remember in SQL you never have the whole DB on the client machine only the part of it that came back with the last query. This is a fundamental difference--it calls for some serious thought about redisigning user interfaces which reliedon having the whole DB on hand for every operation (which is fine for small datasets but impossible for big ones)
Good Luck