On Feb 8, 2009, at 7:14 PM, Antony Blakey wrote:

>
> On 09/02/2009, at 1:08 AM, Damien Katz wrote:
>
>> Nope, each individual read operation gets a snapshot of the  
>> database. When you are replicating, there is one read operation for  
>> every document that must be sent. Each read request gives you a new  
>> MVCC database.
>
> This prompts the question - why not?
>
> It seems to me that the consistency characteristics of replication  
> could be improved if a) MVCC boundaries were created by update  
> operations and b) a given replication was bounded by the latest MVCC  
> state available when the replication was started i.e. replication  
> cannot race with updates. I presume replication cannot use a single  
> MVCC state because it needs to see revisions? If revisions are  
> visible within a state, then my question would be why replication  
> doesn't use a single MVCC state.

It's possible to use MVCC for replication. You'll need to create  
special HTTP command to return you all the documents you are  
interested in a single request, and a special replicator that uses  
that command and loads those documents and writes them to the  
destination.

-Damien

>
>
> Antony Blakey
> -------------
> CTO, Linkuistics Pty Ltd
> Ph: 0438 840 787
>
> Borrow money from pessimists - they don't expect it back.
>  -- Steven Wright
>
>