Consider the geodesic problem on a flat plane. We have to find the extremal function $y$ such that: $$L[y] = \int_{a}^{b} \sqrt{1+(y')^2} \mathrm{d} x$$ is minimized, where $y(a) = y_1, y(b) = y_2$.
Applying the Euler-Lagrange equation to $L(x,y,y')$. $$\frac{\partial L}{\partial y} - \frac{\partial}{\partial x} \left(\frac{\partial L}{\partial y'} \right) = 0$$
When simplifying the above equation, we say that $$\frac{\partial L}{\partial y} = 0$$ as $L$ is not dependent on $y$.
My question is, how is that allowed? If $L$ is a function of $y'$, you will see a $\frac{d y'}{dy}$ term, which is equal to $$\frac{d y'}{dy}=\frac{\frac{dy'}{dx}}{\frac{dy}{dx}} = \frac{y''}{y'}$$
Am I incorrect in saying the above statement? What am I missing?
This is a notation problem.
Write $L(y) = \int_a^b Q(x,y(x), y'(x)) dt $, with $Q(f,g,h) = \sqrt{1+h^2}$.
Then Euler Lagrange says ${\partial Q(x,y(x),y'(x)) \over \partial g} = {d \over dx} H(x)$, where $H(x) = {\partial Q(x,y(x),y'(x)) \over \partial h} $.
Note that ${\partial Q(f,g,h) \over \partial g} = 0$.