i do not see a major issue with the process execution approach. it could be wrapped behind a nicer filesystem-specific API (which is a great idea) but i do not understand the issue you are perceiving