Re: [css3-background] vastly different takes on "blur"

On Jun 11, 2010, at 5:21 PM, Tab Atkins Jr. wrote:

> On Fri, Jun 11, 2010 at 4:58 PM, Brian Manthos <brianman@microsoft.com> wrote:
>> Testability question.
>> 
>> As we converge on the final language, can we make sure that it expresses where -- regardless of specifics of blur algorithm -- a test harness could check for fully transparent or fully opaque pixels?
>> 
>> What I mean is that there should be a no-man's-land beyond the edge of the specified blur region where no blurred pixels should be found.
>> 
>> 
>> My impression is that the intentional flexibility in the specification is (a) about how the blur ramps up / down -- linear, exponential, etc. -- and the impact of neighboring pixels and (b) NOT about allowing the blur to bleed off to infinity or to shrink to barely perceptible.
> 
> Correct.  Defining the outer edges of the blur is precisely what this
> thread is about.  We want to be able to specify that precisely, even
> if we don't specify what happens inside those boundaries.

Yeah, exactly. Well, the inner edge too.

I'd like to be able to say that, at least for parts of the shadow not being influenced by a corner or curve, that a 50px blur means that the inner and outer edges of the blur are no more than 50px apart. In theory, maybe 48px is OK, if the last pixel of the blur rounds to zero% and first pixel rounds to 100%, but blur at 51px is right out.

Even though we don't say what the algorithm is for how quickly the transition occurs, I would hope it wouldn't be something like, say, "each adjacent pixel is half the opacity of the other", as this would leave many of the pixels as good as zero in a wide blurring area. Maybe we can say that anything to the outside of the pre-blurred (straight, not-influenced by curves or corners) edge should be less than 50% opacity, and on the inside must be over 50% opacity. Current implementations seem to do that OK.

Received on Saturday, 12 June 2010 03:52:31 UTC