Haas: Big ideas [for PostgreSQL]

Yeah right. Remember PGCluster? PGCluster II? Postgres-R? pgReplicator? Daffodil Replicator? GridSQL?

As far as I can tell, all of these projects touted to bring multi-node scalability to PostgreSQL, some more more ambitious than others, and some even claimed no drawbacks. Who uses them? If none of these projects managed to solve the problem, why would Postgres-XC?

As far as I can tell, the _only_ clustering and write-scalable system that anyone significant has used, is pgpool-II, and it places significant restrictions on the way you can write your application. Why do people keep thinking that it's a simple problem that can be solved once and for everyone?

Multi-master is just too complex for a database that aims to support ACID properties and all kinds of complex queries -- it actually tends to scale negatively. It's a cost-benefit tradeoff, but if you do outgrow monolithic SQL beasts, then NoSQL and eventual consistency is the way to go. I'll be glad to be proven wrong, however. :)