NoSQL is a regression in a sense. Or, if you prefer, a conscious trade-off. Providing the same guarantees as traditional relational databases becomes difficult

NoSQL is a regression in a sense. Or, if you prefer, a conscious trade-off. Providing the same guarantees as traditional relational databases becomes difficult at scale, so you trade some of the features for higher scalability. For example, you can have a very highly available, distributed key-value store, but with no transactions across multiple rows, no query language, no schema, no indexing, etc. If you think about it, the only real (and most important) advantage of NoSQL databases is that you can make them scale. If only we could make relational databases work like that, I'd happily shove all the data into a store supporting schemas, a powerful query language and ACID transactions.