Once a system is running, it changes with time. Logs are filled, users put their data somewhere, databases are filled. This is not configuration, it is state. I do not want to store state, I want to store configuration. The line between the two is sometimes thin. A system may be configured to periodically fetch and apply available patches to its software. After two years of running and being patched, what is its configuration? Is it the way it was installed originally, or is it the way it was installed originally plus two years' worth of patching?