It is not easy to call somebody’s work garbage, or as Torvalds likes to calls some code “idiotic”, because some people have invested time and intelligence into creating their Magnum Opus. Yet some of that software is definitely wayward much in the fashion of teenage children.
Take for example the latest CouchDB craze. The designers of that product thought that computers do not need to always recalculate the data humans demand out of them, so just pre-calculate everything and present it when it is asked for. Now hold on, the strong point about computers is lightning fast repeated calculations without deviations. The strong point about human and RNA at large memory is quick access to vast amounts of information even lossy sometimes. So we now have a paradigm that we are using the worse of both worlds, a db that remembers calculations. Why bother? especially when it explodes a few megabytes of data into a quazillion of storage. Re-indexing also takes forever and oh OPs are not that bad.
Which leads me to the next major contender, mongodb, which in the paper seem to be OK. Now don’t get me started with the master-slave situation on mongoDB. If you restart the slave with autoresync, it of course resyncs …. from start after purging all its data! As for sharding buh humbug. Only the master has write permission which may not be transferred on a subsequent master election.
Sorry guys and gals, you got it wrong, NoSQL is no panaceia, it will remain a fringe paradigm.