The definition of <size> in ray() is:

Decides the path length used when offset-distance is expressed as a percentage, using the distance to the containing box.

I guess this is pretty similar to containing block, right? So should we choose a different edge based on position property? Or just always use border edge or padding edge? It seems always using padding edge may make sense becuase it looks like contain flag doesn't expect the element overlaps the borders, but I'm not sure.

If the element is an SVG element (e.g. <rect>), how do we define this containing box on SVG layout? Perhaps we need a better definition for this.