> The scenario would be that people would take their laptops away
> from their desks during the day and gather information and then
> return at the end of the day and connect to the lan (by a direct
> network connection using the USB port) and synchronize with the BE
> on the server.
Well, if you have people working on the LAN editing the back end on
the server, I wouldn't have your laptop users synch with the back
end the LAN users are editing. That can cause replication errors
because people are editing records/tables that wouldn't be able to
be synched (depends on the circumstances). Most vulnerable are memo
fields, so those should only be edited unbound.
The best thing is to have a hub replica that the offsite users synch
with, and have that on a synch schedule with the production back
end.
This is most easily implemented with the synchronizer to run
scheduled synchs, but you could code it yourself if you wanted to.
> People may come back to the office
> at the same time and try to update at the same time which I think
[quoted text clipped - 3 lines]
> laptops were synchronized the tables would have different records
> with the same primary key.
Yes, of course, if there are PK collisions, that would be a problem.
But if you're using Autonumbers as PK, that isn't a pragmatic
consideration, as in replicated databases, the Autonumbers are
randomized. In 10 years or so of replication work, I've never seen a
random Autonumber collision.
If you're assigning your own PK sequence, then you've got a major
problem. You have to come up with a way to prevent collissions. The
easiest way is to make each replica's sequence unique to it, either
by pre-allocating ranges of numbers, or by using a 2-column
sequence, one column of which encodes the replica.

Signature
David W. Fenton http://www.dfenton.com/
usenet at dfenton dot com http://www.dfenton.com/DFA/