Why do we seek to minimise $||\vec{w}||^2 / 2$ when the margin for an SVM is $2 / ||\vec{w}||$

Question

According to this article, the following applies to the margin for an SVM: To maximise the margin we must maxmise $2 / ||\vec{w}||$ and to do that one must minimise $\vec{w}$ and then minimise $||\vec{w}||^2 / 2$. Where does the square come from? How is it derived?