> Now that we're using the unified format, I think that this could have been done using either glob or regular expression filtering. Do you prefer sed over those, or is it just a matter of habit?
Habit, I guess. It does produce nicer diffs in unified mode, but that's probably not worth the long-term obfuscation. Let's try and avoid introducing new sed invocations (I don't yet think it's worth cleaning up).