> In reality you will most definitely not write everything under an explicit supervisor. That's the point of the supervisor _tree_. Certainly not every process

> In reality you will most definitely not write everything under an explicit supervisor. That's the point of the supervisor _tree_. Certainly not every process will have its own supervisor, but all processes should be linked to other processes, which are linked to other processors, and at some point you have a process that is quite fundamental to the application and has a supervisor. As long as there's a supervisor somewhere on that tree, the whole subtree will be restarted, hopefully in a non-erroneous state, and the application will continue on its merry way.