It uses a folder with a subfolder for every service. Each subfolder contains a script called run. The system runs the run script. If it exits, it waits two seconds and runs it again. Repeatedly. It's very worse–is–better.
There are commands to control the services and check their status. For example, if a file called down exists next to the run script, it won't run it. This is how you disable a service.
It checks for service folders being created and deleted. New folders are started, and deleted ones are stopped cleanly. They can also be symlinks, so you don't need to worry about deleting a running service folder and you can remove a service from init without erasing the scripts you wrote.
The whole system is useful in many situations and not only as pid 1.
Maybe one day I'll invent a runit–based distribution.
It uses a folder with a subfolder for every service. Each subfolder contains a script called run. The system runs the run script. If it exits, it waits two seconds and runs it again. Repeatedly. It's very worse–is–better.
There are commands to control the services and check their status. For example, if a file called down exists next to the run script, it won't run it. This is how you disable a service.
It checks for service folders being created and deleted. New folders are started, and deleted ones are stopped cleanly. They can also be symlinks, so you don't need to worry about deleting a running service folder and you can remove a service from init without erasing the scripts you wrote.
The whole system is useful in many situations and not only as pid 1.
Maybe one day I'll invent a runit–based distribution.