>Thanks Ed, that sounds like just the ticket. Which solution would you recommend, and why?

I'd choose not to buffer the underlying table; you don't gain anything from having it buffered, and using a double-buffered scenario results in different behaviors when working with local views (which have underlying native tables) and remote views (whose data source resides in a server, and whose underlying tables, assuming the data source is relational, never are opened directly at the workstation.)