The complexity is in locating what is already a distributed database item (git is fundamentally already designed that way), just using a protocol meant for that. Basically just support `git clone` with e.g. IPFS URIs.