Not understanding something fundamental about the Metropolis-Hastings algorithm

I am reading Bishop (which is slightly over my head currently) and looking at the code hosted here, and I believe I understand everything about this implementation. It seems to be approximating a Gaussian distribution, but it looks like it's using a Gaussian PDF to approximate the Gaussian distribution, which is where I get lost. I do not understand why this is beneficial or helpful. Perhaps the implementation is not really all that helpful as the result of being simplified for the sake of demonstrating the algorithm clearly? Anyway, if someone could provide some intuition behind the algorithm, I would be grateful.