Writing one is not a waste of time, because you learn what's really going on just after POST. But trying to actually use your own and maintain it is pretty much a waste. Write one to learn how it works, then throw it away and use Grub.
If you consider that valuable knowledge, then I agree with you there. There may be some merit to writing (and understanding) a minimal bootloader that enables A20, switches to protected mode, and loads a kernel from a floppy disc. Just getting that far will teach you quite a bit about writing and debugging low-level assembly code. However, if you go this route then, as you say, you should definitely consider it throwaway code.
But, I don't think that knowledge is important. You'll gain much of the same knowledge from implementing virtual memory and multitasking. Besides, the bootloading process is highly architecture-specific. Switch to an ARM or PPC chip, and it's a completely different process.
Switch to an ARM or PPC chip, and it's a completely
different process.
Yup. That's why you should do it for learning purposes - getting your head around how a machine bootstraps itself is, to my mind, one of the harder things to understand about a computing system. Part of that is because someone who doesn't know how it works might look at booting mechanisms as a sort of "black box", or just magic. Once you peer into the black box and understand some of it, suddenly, none of it seems all that mysterious anymore, even if you've only learned the particulars for one specific machine.
