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

On Sat, Jun 12, 2010 at 3:38 PM, fantasai <fantasai.lists@inkedblade.net> wrote:
> On 06/12/2010 12:05 PM, Brad Kemper wrote:
>>
>> On Jun 12, 2010, at 11:03 AM, fantasai wrote:
>>
>>> Ok, I've updated the spec text:
>>>  http://dev.w3.org/csswg/css3-background/#the-box-shadow
>>>
>>>  http://dev.w3.org/cvsweb/csswg/css3-background/Overview.src.html.diff?r1=1.230&r2=1.231&f=h
>>>
>>> Let me know if this is better.
>>
>> Yes, the last paragraph you added looks very good to me.
>>
>> I still think the word radius and radii should be removed, except when
>> talking
>> about border-radius, but instead it seems to be there even more. The way
>> we
>> are defining it, it is not a direct input of the Gaussian blur (Dave
>> Benning
>> seemed hopeful that it could be mathified into such an input, to give the
>> results [or perhaps close approximation?] of the last paragraph added).
>> and
>> it does not seem to be the radius of any other circle... was this a
>> regression
>> of your earlier text where you had mostly removed that term?
>
> Fixed.
>
>> I also start losing track of where the sentence is going in the part about
>> positive and negative number, absolute values, etc. Maybe that part would
>> be better broken into bullet points for more clarity. A negative spread
>> on an outer shadow has the same sort of effect on the shape as a positive
>> spread on an inner shadow, and vice versa, but I didn't really get that
>> sense while reading this.
>
> The general sense of what happens is given above, where the values are
> defined. This paragraph is the details of exactly how to interpret it.
> Anyway, I've tried to fix up the paragraph a bit, take a look now.
>
>> When you say "a positive blur [radius]", it sort of implies that there
>> could
>> be a negative blur. So I think "a non-zero blur amount" would be better
>> there.
>
> Fixed.
>
> ~fantasai
>

To me, the new text actually seems more "you know what I
mean"/"hand-waving" than before.

This might not be tenable, but thinking of Simon's SVG filter
suggestion and looking at the spec for feGaussianBlur
http://www.w3.org/TR/SVG/filters.html#feGaussianBlur
this whole conversation might be made a lot simpler by simply
specifying the blur used rather than leaving it undefined.

If the shadow is blurred by a Gaussian blur (or its box filter
approximation) with a standard deviation of the blur-radius divided by
3 (or the "blur width" divided by 6), the effect will be exactly what
we all seem to mean while still leaving room for efficient
implementations. Based on what Rob posted, this is what Gecko is
already doing.

This leaves the problem of defining the allowable approximation
without a dependency on SVG, but since AFAIK all the major UAs already
support this effect (either through SVG filters or IE's Blur filter),
it seems like there should be a simple way to do this.

Received on Monday, 14 June 2010 21:57:08 UTC