Why build it to be a drop-in replacement right now? You don't even know what kind of datastructures you're going to have. You're overthinking this. Literally just get something so stupidly basic it sounds like it won't be useful for anything out there right now. Then get the next version out there 7 days later. And the next 7 days after that. And so on and so forth. Linux 0.01 didn't have a windowing system - it barely booted up an x86 processor.
My criteria for good software:
- Does it work?
- Can someone else come in, make a change, and be reasonably certain no bugs were introduced?