>My guess is that network buffering issue is very limited in the time, in other words, need to catch the moment very precisiously when to switch off the power to catch the moment when data updated from VFP and not saved because buffering in the network and at the server machine.

Right, the last time I checked this (wish I could remember where it was) I set it to one second. But then, I don't remember if we still had problems after that or not.

>>Sometimes the files were coming in by modem, sometimes by floppy, but the general replication scheme worked, based on each station's prefix on primary keys.
>
>
>VERY nice solution, though some data still required to be as latest as possible (real-time).

In this case this was the best we could do. They'd of course have liked it if they could have real-time data, but they were simply not equipped to do it - some of the scales were in a middle of a field, and even had problems with mice. The mice actually destroyed one printer :).

> In such case it will not work and other solutions are required, sch as mirror data (writing into the 2 databases) that will assure at least one of the database after power off contains correct record.

Under my scheme, it would require synchronization to take place at sufficiently short intervals, or on demand. If all the workstations are already wired together (i.e. visible on the same network), that'd be feasible.