Evans page 250&251
He gives us two theorems. Local approximation by smooth functions and global approximation by smooth functions. The second seems to give the first without any additional assumptions, what gives?
The global version says that $U$ is bounded, and the local does not say this. But immediately the local uses $U_\epsilon$ which is defined via the boundary of $U$ so evidently $U$ is bounded?
Then the only other difference seems to be that the approximating functions are $C^\infty$ on $U_\epsilon$ rather than on $U$. Why do we care about that? What do we gain by having these approximating functions defined on a greater region?