The evolving docker ecosystem vs the devolving hadoop ecosystem.

I had a chance to learn about the way RunC is changing things today from dan walsh at red hat.

In kubernetes, we directly depend on rkt/docker/etc APIs for doing things, but with RunC, an abstraction exists which does all the things we care about (Dockerfiles, microservice start/stop, etc) without being coupled to any individual vendor.

Currently, The ContainerD RunC border is a little complex to grok IMO.

But in the end, the evolution of the docker processes into a completely wide open OCI based set of applications is really cool - it basically means docker, and kubernetes, will be able to continue evolving above the stack.

My only question: Why didn't this happen with hadoop a long time ago ?  It seems like the bigdata vendors, on the other hand, are on a race to the bottom.