After staring at the MatLab[https://www.mathworks.com/help/matlab/ref/atan2.html#buct8h0...] explanation, I have to admit I don't really get that. Minus zero is a encoding artifact. Values of [x,y] with either `x == 0` or `y == 0` don't have, in the mathematical sense, a quadrant. The fact that a convention preserves an encoding artifact doesn't make it useful.
E.g. I've seen it argued that +0 and -0 represent limits of functions that approach zero from the negative side or the positive side. But that makes no sense, because no other floating point number represents a limit going from above or below the number, so why should 0? Also, all this walking on eggshells around a possible -0 (because that represents the limit of 0 coming from below 0) is just lost when someone writes 0. Admittedly, I've not seen the guts of most numerical code, but the only time I have ever seen a minus 0 written into code was to test the minus-zero handling of something.
