Actually, relational databases are the relative newcomers. The limitations of things like CODASYL were well known to 1960s programmers and Codd's relational model sought to address them.
Why the non-relational databases keep getting reinvented has been a bit of a mystery when there is already a rich history of development to look to. I get the feeling a lot of the industry isn't much into history, especially of pre-micro computer systems.
Network databases, key-value stores, graph databases, commercial offerings like Pick... you'd think the NoSQL people would have looked into it all before proclaiming their new found solutions, but apparently not.
