On Aug 25, 2009, at 9:51 AM, Robert O'Callahan wrote:
> On Tue, Aug 25, 2009 at 8:55 AM, Tab Atkins Jr.
> <jackalmage@gmail.com> wrote:
> Indeed. As I currently have it, and as Webkit does it, none of the
> background-* properties leak into the gradient() rule. It's a
> completely independent function that just does its job drawing an
> image with the information provided and then steps back, just like
> url().
>
> The logic here is that repeating behaviour depends on the image type.
>
> I'm inclined to keep it that way. To do otherwise is to make it act
> unpredictably when further combinations of background-* properties are
> used. For example, when you use background-size to make the image
> less than the size of the box, what happens? Does it repeat?
>
> No, it scales. Whether an image repeats or not is independent of the
> image type. The only thing that varies by image type is how it
> repeats.
>
> What if
> you want a gradient that only runs through its stops once (which would
> require no-repeat), but is then tiled across the background (which
> would require repeat)?
>
> You basically never want that.
I downloaded the Firefox nightly to see what ti was doing. Here's a
screenshot (attached, if you don't see it inline):
