TTY-based group scheduling

- I have always been told that kernel should implement mechanism and not policy and this sounds like a policy to me. Not sure where to draw the line
though.

- tty groups are hidden and and don't appear in cgroup hierarchy and are not user visible. So if some controller is implementing the upper limit, these groups might not be subjected to those limits.

- Once cgroups are visible to user space, one can write a nice monitoring tool to list all the cgroups, their resource consumption, processes running in these etc. With hidden group, one can't do that.

So IMHO, this kind of automatic group creation is more of policy and should be done in user space. Even if we do it in kernel, atleast we need to make sure these groups become visible in appropriate cgroup hierarchy and are user controllable as regular cgroups are.