Time to reject traditional database techniques? (no!)
I think pundits crying the demise of relational DBs frequently miss the point—the relational abstraction will always be a useful one, the problem is that it’s harder to scale databases at that abstracted level.
So tools are arising which scale the DB at a lower level of abstraction, leaving you to re-implement relational techniques poorly ontop of them.
Instead I would like it if someone actually tackled the really hard problem—elegant and efficient distributed implementations of relational algebra.
Oracle have tried and could probably get a lot more traction on it if they put their minds to it—but so much legacy SQL cruft is entrenched into the ‘enterprise’ that that may be hard.
