News for hackers. Stuff that matters.
Re:Why Replication is Awesome by memracom on Thursday November 07, 2013 @02:03PM (#6691613)
Meh... Replication has been around for a long time. Couchdb was not anywhere near the first, in fact Couchdb was just leveraging capabilities that were in Erlang OTP for many years.

And replication is not perfect either. There are tradeoffs and corner cases that need to be understood. Riak is another cluster datastore that is written in Erlang OTP, like Couchdb, and the people who build Riak are quite open about the issues that they have to deal with and the corner cases. They often present at conferences and write blogs on these topics so if you really want to UNDERSTAND replication, google for blogs and conference presentations connected to Riak.

In any case, Couchdb is good, Riak is good, but even traditional RDBMSes like PostgreSQL are good and can do replication. In all these cases, the developers have wrestled with the math and computer science behind replication and have made something that mere mortals can use.

Re:Why Replication is Awesome by BigBlueHat on Thursday November 07, 2013 @03:28PM (#6692157)
To be clear, "replication" in this case is "multi-master replication."

Riak's Replication: http://docs.basho.com/riak/1.3.2/references/appendices/conce...

CouchDB's Replication: http://docs.couchdb.org/en/latest/replication/intro.html

Riak's is for data safety within the cluster.

CouchDB's is multi-master, cross-cluster, cross-device, etc. It can either be uni-directional (push or pull) or (if you do both) it can serve to synchronize two distributed master databases--each serving as a primary write point in the architecture.

That seems unique.

Re:Why Replication is Awesome by rdtsc on Thursday November 07, 2013 @02:23PM (#6691766)
> Replication has been around for a long time.

Well yeah data copying has been around as long as there is data. Peer-to-peer replication in the database world, with consistently and openly defined protocol has not.

http://dataprotocols.org/couchdb-replication/

I haven't seen too many databases that do this well. And well enough to have other database product interface with them (PouchDB for example).

> They often present at conferences and write blogs on these topics so if you really want to UNDERSTAND replication, google for blogs and conference presentations connected to Riak.

I don't remember having too many issues. What exact issues did you encounter with CouchDB's replication? You emphasize UNDERSTAND, do you have any indication CouchDB developers don't UNDERSTAND replication?

Re:Why Replication is Awesome by mlmilleratmit on Thursday November 07, 2013 @05:28PM (#6692883)
Sure, nobody claims that CouchDB is the first to have replication. MVCC was invented decades back, and hash trees have been around forever. I believe those two combined are compelling, and if you follow, e.g. risk, you'll see they're heading down the same path.
Re:Why Replication is Awesome by patrickmay on Thursday November 07, 2013 @02:11PM (#6691679)
> Replication has been around for a long time. Couchdb was not anywhere near the first

Indeed. I occasionally work with GigaSpaces -- they had replication at least five years before CouchDB existed. And they had competitors before then, too.