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

On Tue, Jul 27, 2010 at 5:50 PM, Brad Kemper <brad.kemper@gmail.com> wrote:
> On Jul 27, 2010, at 2:57 PM, "Tab Atkins Jr." <jackalmage@gmail.com> wrote:
>
>> Specifying just the point at which it must hit 2% or whatnot allows
>> things like a linear gradient being used as a shadow, which we'd like
>> to avoid.
>
> I'm not convinced yet that _we_ shouldn't allow that (well, a shape gradient anyway, as linear gradients only go in one direction). I already bought into the part of the spec that says a specific algorithm is not required, and you haven't yet said anything to change my mind.
>
> If Gaussian is truly the no-brainer you say, then all implementors will use it or a close enough approximation anyway, with the same results. But I don't see why a UA shouldn't be allowed to create a nice looking blur in some other manner, such as a gradient between two corner softened shapes (if it's ugly, the implementors will hear about it, or users will move to prettier UAs), or why they shouldn't be allowed to blur by other means or maybe put some jitter or noise into the blur.

The basic reasoning is that, if we know what we want and that
implementations can do it, why *not* tie it down tightly enough to
actually guarantee that an impl labeled "conformant" is doing what we
want?

Inserting looseness into a specification "just because" doesn't seem
very good for anyone - impls aren't instructed on the "correct" way to
do things, authors can't rely on impls necessarily doing the "correct"
thing, and tests aren't as useful.  I only supported higher degrees of
looseness previously because I thought we needed it to capture what
impls were willing to implement.

~TJ

Received on Wednesday, 28 July 2010 01:09:18 UTC