@koro About 2-dimensional problem of this, $\{x\in\mathbb{R}^2:\text{exactly one of }x_1,x_2\text{ is rational}\}$, then this is easily seen to be totally disconnected since its complement contains $(r_1, r_2)+(\pm x, \pm x)$ where $r_i$ are any rationals and $x\in\mathbb{R}$
But this argument doesn't quite work for $\{x\in\mathbb{R}^3 : \text{exactly one of }x_1, x_2, x_3\text{ is rational}\}$
Because such lines don't enclose a 2-dimensional shape
Is this solution correct? I've looked up the solutions to this problem and every solution uses more complex approaches so I'm unsure if there is some basic flaw in this solution that I'm missing out on.
you write b <= a, c <= a, d <= a in succession (where a = cos(x) and b = a^4 and c = a^2 and d = a), then you write just below that, "implies" b - c + d <= a. what is going on with the "implies"?
@Koro in general, this question is not quite well-posed because $M$ is not a subset of $E$. the LES in homotopy implies that a necessary condition for $\pi$ to be a homotopy-equivalence is that the fiber $F$ is weakly contractible (hence contractible under mild technical assumptions). if the fiber $F$ is contractible, it turns out (under mild technical assumptions) that $\pi$ has a section which turns $M$ into a deformation retract of $E$.
(in particular, this is true for vector bundles, where the proof is of course much easier)
@Koro yeah, that's a nice argument
another argument that uses cohomology instead of homotopy, but is more convenient than the one you've screenshotted, is that if $\pi$ were null-homotopic, you would get homotopy-equivalent attaching spaces $\mathbb{CP}^{n+1}\simeq\mathbb{CP}^n\lor S^{2n+2}$, but these have non-isomorphic cohomology rings