Decoupling Plugins from the real hadoop-cluster via Cluster-Proxy

In order to decouple the eclipse plugins from real clusters,
I would suggest to work with web services on the developers machine,
which connects to real clusters and act like a command gateway.

All plugins act just as consumer of the web services, which map to real cluster services.
Each cluster proxy hast its own configuration. So we can compile against different versions
and even connect to multiple cluster. Libraries are managed via maven dependencies and
the plugin structure is much more easy without the need for playing around with multiple
hadoop versions.

Would this be a reasonable approach?