Playbahnosh:
All in all, today's physics engines are a joke, nothing more, only there for spectacle, but have very little to do with actual physics. If Irvin and Smith could make a simulation that accurate, that could fit on one side of an audio casette, why can't today's developers make a real physics engine?

While there are still some technical answers to that question (3D is way harder than 2D; spare computational power is going to better graphics, not more accurate simulations (and that often yields a better return in terms of improved immersion); thermodynamics and fluid dynamics and so on are much harder than mechanics), you're right that "physics engines" are often movie physics engines. But what's wrong with that?

Of course, you want the physics in a game set in the real world to be somewhat realistic, that avoids jarring bits of unreality or mismatches between player expectations and game mechanics. But players also expect a degree of unreality. (In an action game, I'd expect a thrown barrel to fall in a parabola, but I'd expect bullets to take a straight-line path, no matter the distance.) Furthermore, the complexity realism adds can add nuance or distraction. (Having heavy items weigh down your jetpack is nifty, but if that requires you to spend much of the game micro-managing your inventory, that will detract from most players' fun.)

One game I'm looking forward to is Just Cause 2, which boasts a physics engine that intentionally deviates from realism. Would that game be more enjoyable if the physics engine was more realistic? I doubt it.