The commands don't require you to know the underlying storage model. e.g. I have maybe once in 15 years had an occasion to think about packfiles. I couldn't tell you basic facts about the storage like whether a git clone will mean you necessarily have the same object files as the upstream (I presume no). Content addressed (with named references) snapshots aren't the underlying storage model. They are the UI model, and they make a ton of sense for that purpose.