[michaelmior]

For CSV, "breaking" changes seem like less of a big deal to me. Partially because there is already so much variation in how CSV is implemented.