No.

I think you just needed more discrete mathematics and more combinatorics and graph theory. Then the compiler course would've made a lot more sense up front when learning about recursive-descent parsers vs. LL(k) parsers vs. LR/LALR parsers and where the strength of each one lies, how it's making your life easier/harder, etc. I did that middleware OS stuff you mentioned though, and I have to agree that it was sometimes pretty useless. Working directly with, say, a Linux, FreeBSD, or Minix kernel+OS would've gone a lot further and made you feel like you were accomplishing something that could immediately be put to use. I agree with the idea of pushing/encouraging your students to learn research on their own and to learn how to WANT to research. Bjarne was also advocating that in the summary (I didn't read TFA), and I think it's the most important step to teaching.

Actually, you should look up "dependent". You say we it's okay to use it, but we shouldn't be dependent on it. Well, then why use it at all? Dependent just means we require its use, and if we want more accuracy, we do require its use. I'm preeeeeeettty sure we will still know how to use and will still have radar in case there are problems. You know, that same government you mention could very well do the same thing with our existing radar network. So, what's the difference? Also, with satellites, we can still do visual recognition if we think something is amiss. So, technically, we're getting even MORE failsafes. Take off your tinfoil hat now.