On 13 Aug 2015, at 5:28pm, Tim Streater <tim at clothears.org.uk> wrote:
> Presumably all the OP needs to do (in the future, at any rate) is: > > open first db > attach second db > start transaction > do updates to first db; > do identical updates to second db > commit transaction > close connection > > or some facsimile thereof. One 'solution' to this problem is just to take any node's copy of the database, execute all other node's changes against it, then to distribute this copy to each node. This increases the number of bits that need to get from one node to another, but means that you don't need to do lots of processing on all nodes. It's not a good solution and it's easy to find problems with it. But it can be done without defining a 'master node' and without having to design a client/server solution. And, for most installations, where (for example) only one salesman works on each account, it works fine. Simon.
