Andrew Fedoniouk <news@terrainformatica.com> wrote:
> div.case10
> {
> width:10%;
> height:60px;
> border:30px solid;
> border-radius: 60px/20px;
> }
> ?
>
> I am getting this (content updated):
> http://www.terrainformatica.com/w3/round-corners-sciter.png
>
> Cool of course but is the right way of doing this?
What we want to be saying, for the inner-edge-is-a-sharp-corner case,
is something like this: draw lines outward from the inner corner,
continuing the line segments of the inner edge, until they intersect
the outer edge. The gradient must be inside the region so defined.
I think this is what the current wording means, but it *is* confusing,
and my revision isn't the right fix -- as this example shows, the
transition might *not* be supposed to cover all of the area with a
curved outer border.
I've added another case to my diagram.
> >> I believe that the only reasonable type of gradient here is so
> >> called conic gradient. Probably it makes sense to define just
> >> that?
> >
> > I'd call that out of scope for Level 3. There are several shipping
> > implementations that do sharp transitions.
>
> I am not asking to remove optionality of gradient.
Sorry, I misunderstood you. I agree that a conical gradient is
appropriate in the abstract, but I do not want to write an exact
mathematical description of the gradient to use into the spec,
there isn't a convenient other spec to refer to (neither SVG nor PDF
appears to define conical gradients), and it doesn't seem fair to
implementers to give them neither of those.
zw
