NoSQL really is about having no SQL
There's been an effort to change the meaning of "NoSQL" from meaning "no SQL" to "not only SQL". But after using Neo4J for a while and experimenting with other NoSQL databases, I've come to believe a strong driving force behind NoSQL adoption is not scalability and data volumes but the difficulty of maintaining SQL. Programming languages & environments are now much more productive than those of the '70s, when the relational model was invented, thanks to improved compilers and runtimes, language-aware editors that put massive program transformations at one's fingertips and the wide adoption of test automation supported by convenient test frameworks. It's easier to write code that copes with different document structures or transforms documents into new forms than it is to write SQL that migrates a complicated schema, especially when views, triggers and stored procs are involved. All the recent research into making ACID, SQL RDBMSs that can store petabytes of data is missing an important point: the lack of SQL really is a killer feature.