Hurrumph, says the guy who did databases in the 60s. I believe that going with NoSQL is generally a mistake when you want any kind of sorting of human-generated data. Why? Because humans spell things inconsistently. Let's say you want to look for contacts in a given city. In an RDB, you make tables of countries, provinces, cities and you let people pick. With NoSQL, you let save what people hand you. Ashville? Pittsberg? Wooster? Or say you're in India and the government changes the names of cities or states - then what? This leads to people trying to do relational things with NoSQL. I think there are many ways to solve issues of scaling and distributed data while sticking with SQL