I've been thinking about user-space code vs. kernel-space code, and it occurs to me that if you abstracted away the differences between them you could write code that could be compiled for or run in either. And it would be at least as slow as user-space code and would hang the machine in no time flat.
Now, if I only had some decent library access and a clue, I could check this theory against the literature.
Yes, I'm spending my sabbatical navel-gazing. My CS degree has prepared me for this wonderfully.