On Mar 8, 2009, at 9:39 AM, Chris Anderson wrote:

> The second is that you have to think this way if you'll be dealing
> with replicated data. If you have edits on dbB and you replicate them
> to dbA, dbA only see the edits that dbB would still contain if it were
> replicated.

It is possible to preserve version history in a distributed system,  
though. Git, Mercurial, Bazaar etc. all do it. I imagine the same  
techniques they use could be implemented at the application level in  
CouchDB, although it seems a bit inelegant to then have two different  
notions of versioning on top of each other.

The "Revlog" structure used by Mercurial for this is quite elegant (it  
seems to have a lot in common with CouchDB's implementation!) and is  
nicely explained in this paper [PDF].

—Jens