RE: Bi-directional transactional replication on IBM i

We have a few files that we replicate across 3 systems. All three update the files. We basically wrote a module to perform all the files updates. We use before and after images from a trigger to do field level updates. We chain to the record w/lock and compare the before to after at the field level, if the field changed, compare the before to the current, if the same we update. If Programming logic appears to be complex but it is one set of code repeated for each field. Makes heavy use of externally described data structures, b.field = before, a.field = after, and plain field = current. You can log any conflicts and update the field with the new value anyway discard the update to the field. Which ever you choose. It does require you to send both the before and after images to the remote system(s). For the most part, we do not see the same record being update on multiple system before the previous updates have been applied. We also use data queues to store and updates until the remote system can read the data queue.