Instead of using kubectl and docker CLIs, switch to compiling against the official APIs instead. This will eliminate the prerequisites and at least in the case of Kubernetes, increase performance as (I assume) we'd be able to reuse a single or small set of TCP connections for each request rather than calling kubectl each time which will create new TCP connections to the API server for every request. It will also reduce the headache associated with leaky processes which is particularly annoying on Windows (and really not implemented at this point).
Instead of using
kubectlanddockerCLIs, switch to compiling against the official APIs instead. This will eliminate the prerequisites and at least in the case of Kubernetes, increase performance as (I assume) we'd be able to reuse a single or small set of TCP connections for each request rather than callingkubectleach time which will create new TCP connections to the API server for every request. It will also reduce the headache associated with leaky processes which is particularly annoying on Windows (and really not implemented at this point).