Talk:GEPS 009: Import Export Merge
Perhaps, instead of inventing our own synchronization protocols, we can hitch a ride on something ready-made for the job -- with the current state of data (with to_struct), or with UIDs added where needed to data?
I have an intuitive feeling that git might help -- because it is a tool with multi-headed partial synchronization in mind, and because it has built-in mechanisms for locating duplicates of the same data. We'd need to map our data onto git plumbing so that it's possible to diff at our objects' field level, and to make nice addressing of the objects involved (take this with a grain of salt -- I've only used git porcelain, never done deep git hacking). Or maybe use something else in-between (smth like git-annex)? Just imagine operations like "pull Doug's changes to such-and-such family and their descendants!" with all the edits to the same people auto-merged when there's no conflict! BACbKA (talk) 14:08, 30 September 2013 (MST)