On Mon, May 25, 2009 at 01:44, Scott Shumaker <[email protected]> wrote:

> If that's the case, it does explain why there are some decisions in
> CouchDB I find strange, like the fact that authentication runs during
> replication.  You're using replication for a completely different
> purpose than I need it for - I need it for redundancy and read
> scaling, you're using it to synchronize disparate data sources.  Very
> different problems that call for very different solutions.


I called out this distinction between usages in my message above.
If the replication functionality were "directed" or overseen by the
sharding/partitioning code in the data center deployment, I think we can
totally prevent replication conflicts *caused by* the partitioning code.
However, replicating to/from the data center could still create conflicts in
the normal way.

Reply via email to