The only thing you need to make replication work are the algorithms around couch_key_tree. Then if you want to make it incremental, you need the update sequence index. After that its all just optimization.
On Thu, Sep 1, 2011 at 9:00 AM, Benjamin Young <[email protected]> wrote: > Oh dear... ;) Not at all the same thing. :) > > I agree with Stuart that it'd be great to see the CouchDB replication > "protocol" (HTTP + MVCC + _changes feed...mostly) codified into a > specification that could be implemented anywhere. There's working being done > to prove that it can be with Mikeal's PouchDB project: > https://github.com/mikeal/pouchdb > > Perhaps the first step looks like listing off the key pieces of the puzzle > and things that might be less obvious like needing consistent, though > arbitrary, conflict "resolution" (always selecting the same winner). > > Anyone know if there's a wiki page or blog post that would get this process > started? If not, anyone know enough about the feature to start one? :) > > Thanks, > Benjamin > > On 9/1/11 9:07 AM, Roger Rohrbach wrote: >> >> This isn't exactly what you're asking for, but behold: SyncML >> (http://www.openmobilealliance.org/tech/affiliates/syncml/syncmlindex.html). >> >> And weep. :-( >> >> >> On Sep 1, 2011, at 3:09 AM, Stuart Rackham wrote: >> >>> In the same way that UnQL standardizes document database queries does >>> anyone know of any work being done on a standardized document database >>> synchronization protocol? > >
