Second part was showing that the converse was true. Now, the path that Dummit and Foote takes is basically saying aight, you can write $L = K(\sqrt{a+b\sqrt{D}})$ (where $K = \mathbb{Q}(\sqrt{D})$, so show that $\mathbb{Q}(\sqrt{a^2 - b^2D})$ is another intermediate extension so it must be equal to $\mathbb{Q}(\sqrt{D})$.
Now we had an earlier exercise that section saying that if $F$ is a field containing $n^{th}$ roots of unity (characteristic not dividing $n$), then $F(\sqrt[n]{a}) = F(\sqrt[n]{b}) \iff a = b^ic^n$ for some $c\in F$
And they were like, okay use that to conclude that $D$ can be written as a sum of squares
In the process of doing that I computed that $\alpha = \sqrt{a+b\sqrt{D}}$ had minimal polynomial $x^4 - 2ax^2 + a^2 - b^2D$
This was in order to say that $\sqrt{a-b\sqrt{D}}$ was another element in our field, and so the Galois automorphism taking $\alpha$ to this (order 4 since we're $\mathbb{Z}/4$ and there's only one element of order 2 already given by $\alpha\mapsto -\alpha$) cannot fix $\sqrt{a^2 - b^2D}$, so that can't be in $\mathbb{Q}$