Re: [css3-background] Where we are with Blur value discussion

On Mon, Jul 26, 2010 at 9:41 AM, Brad Kemper <> wrote:
> On Jul 24, 2010, at 2:47 PM, Dennis Amrouche wrote:
>>>> What makes blurs special such that the extent parameter is only interesting at < 300px?
>>> The fact that they're blurry, mainly.  And they're not really used for
>>> precise layout purposes, so a shadow blurring to insignificance 10px
>>> shorter or longer than you expect when it's extending 300 or 400px out
>>> already is almost certainly unimportant.
>>> ~TJ
>> I heavily disagree.
>> An attribute and its value is exactly to print into the browser window what they express to the author.
>> Models should be as complete as possible, specs should be designed and implementations done as tight as possible here.
> I think the issue is that at some point, when describing how far out the shadow should extend, you have to define what is part of the shadow and what is not. Thus, 10% opacity is part of the shadow, but < 1/256 opacity probably isn't (since a 24bit photo editor can't even detect the difference from totally transparent at that point). At the blur gets bigger, it becomes less obvious where this distinction should be made. Is less than 1% opacity still part of the shadow that should be measured for our purposes? If we base it on my personal human perception, then it seemed like anything less than about 3% was pretty darned invisible.
> So if the implementors (such as Tab, now) are saying that it should be around 2% opacity, or even 2.4% opacity or whatever, in order to keep the math manageable for better performance, then I'm cool with that, but I'd like this point to be consistent across UAs. I'm kind of a purist, but I have to admit that if the actual results are a pixel or two off from 200px, not even the most finicky designer is likely to notice.


Unlike a shadow spread where, if you used an opaque shadow color, the
edge is bright and sharp and visible, on a blur the "edge" is
impossible to accurately detect without tooling.  It's just
"approximately the point at which the shadow becomes invisible to the
naked eye".  That covers a range of distances in the first place, so
it's an intrinsically fuzzy number.  I only picked 2% because I
stumbled onto the fact that it has a particularly nice relationship to
the stdev of a gaussian blur.

So, yeah, like Brad says, if the "range of pixels within which the
shadow becomes effectively invisible to the human eye" is a bit off at
very high shadow blur lengths, you'll never notice.  It's a pretty
unimportant point.


Received on Monday, 26 July 2010 16:59:58 UTC