Neither of those options would actually help anything though. At least with df, you'd either have the parsing group opting into the non parsing group if it's faster, or vice versa, and we'll just be in the same place we are currently.