There are two ways of storing a color with alpha. The first is exactly as you see it, with each component as-is. The second is to use pre-multiplied alpha, where the color values are multiplied by the alpha after converting it to the range 0.0-1.0; this is done to make compositing easier. Ordinarily you shouldn't notice or care which way is implemented by any particular engine, but there are corner cases where you might, for example if you tried to increase the opacity of the color. If you use rgba(0, 0, 0, 0)
you are less likely to to see a difference between the two approaches.