So you're saying the problem that this article is solving is just preventing the language from generating code that can possibly load some incorrect bytes from main memory, even though these incorrect bytes will in every case be ignored? Why is this considered a problem at all?