Digging in the kernel dust

By mitigating the impact of wrong decisions.

One of the more interesting features in Erlang is the "let if fail" design ideology. Instead of trying to recover from error conditions you just fail cleanly and restart the affected parts. And then you design your system with the assumption that individual components will fail, except maybe for a very limited minimal subset.

It translates well into microkernel architectures. In theory, that is.