The motivation is that both kinds of "on-premise" create roughly similar engineering challenges. (Your app has to be adaptable to heterogenous execution environ

The motivation is that both kinds of "on-premise" create roughly similar engineering challenges. (Your app has to be adaptable to heterogenous execution environments, all your code changes will be in a partially released state for indefinite periods of time, it's hard to immediately diagnose whether a reported problem is your fault or user error...)