>Hi,
>
>I am wondering what is the best way to define indexes for distributed data that permit to synchronize it with central data without index collisions.
>
>One posibility is that all pk's be composed of two fields: location_id + autoincrementing_id, another is to use GUIDs.
I once used location_id c(2) + subs(sys(2015),2) as a PK, and that was in FPD2.6. The data were exchanged as zipped dbfs, via modem. Worked perfectly, except that the phone lines were so bad that the customer gave up. There was a reserve version which would have operated via floppy, but these were wheat mills, with lots of dust in the air, so that didn't work either.