DNS has no git . I built one. This is a working prototype, not a finished product. The point of the post is to show that a long-standing gap in how DNS handles its own state is real, and that closing it is more tractable than it looks. The 2 a.m. question Pick any on-call channel. Wait long enough. Eventually someone types: "What did api.checkout.acme.com point to ten minutes ago?" There is no good answer. You grep five log files. You SSH into the secondary. You hope someone's terminal scrollback is still alive. You make a guess, ship the rollback, and write a postmortem with the words "lessons learned" in it for the fourth time this quarter. We have had a working model for this since 2005. Linus called it git . Twenty-one years later the most critical name-resolution layer on the internet still treats its state as a file you overwrite. No history. No diff. No blame. No rollback. No preview. That gap is what this project is about.…