We’ve known since the 1950s that our configurations should be declarative, to make them resilient to necessary changes to our software stack.
Instead of coding exactly what change needs made, we ought to write a config that declares the intended outcome, and then do extra work to write code that correctly interprets that config. This way when all the commands we used stop working (and they do!), we still know the original intent of the configuration.
But making config management declarative is a lot of work. So fuck that noise. I’ll do it in bash, instead, again.