how is it different from the "sidewalk" algo in the article?

you'd get the same pitfall of torus topology with 2 perpendicular loops