When we branched branch-1 one of the goals was the ability to remove hadoop1 support from master. I propose to do this softly at first by removing it from the poms removing the 20S implementation of the shims.
I am not going to remove the shim layer. That would be much more disruptive. Also, I haven't done the homework to see if we could, as there may still be incompatibility issues between various versions of hadoop2.