Explain briefly, but rigorously, why running Kruskal's algorithm with the squares of the weights yields an MST on the original network (with the original weights). Assume all weights are nonnegative.