Re: [CSS3-images] premultiplied alpha (was rotate(<angle>[, <translation-value>, <translation-value>]))

On Tue, Feb 21, 2012 at 11:56 AM, Brian Manthos <brianman@microsoft.com>wrote:

> Rik:
> > Well, you *could* special case 'transparent' to say use a transparent
> version
> > of the previous or next color stop which is really what the author means.
>
> This was also discussed in depth and had concerns.
>
> Some of the concerns:
> - "special case of 'transparent'": animations and transitions become
> incredibly complex with no "correct" answer in at least some situations
> - "use the transparent version of the previous or next stop": well which
> one?  Or is it both depending on situation?
>

it would be both (unless it's the first or last)

- "what the author means": any spec that implies the editor knows what the
> author means is usually wrong in its generic boldness
>

Well, in the case of going from rgb(255, 0, 0) to rgb(0, 255, 0, 0), I'm
pretty sure the author didn't mean that you should ignore the red value...


>
> +1 @ my posts about improving "memory" coming out of this forum.
>
>
>
> L. David Baron:
> > Choosing premultiplied means optimizing for the normal case rather
> > than the odd feature of transition to transparent + color.
>
> Rik:
> > I don't think going from red to transparent yellow is an odd case at
> all. It
> > gives you a similar effect to InDesign's gradient feather.
>
> I think it's a reasonable scenario but also reasonable for CSS to draw the
> line somewhere regarding what to support directly.
>
> The PROs and CONs of { (a) premultiplied, (b) not premultiplied, and (c)
> both available } were discussed in depth multiple times and the conclusion
> (to my recollection CSSWG resolution) was to go with (a) for CSS3.  There
> was some gnashing of teeth in response, but mostly localized to a few
> voices.
>
>
If this issue was brought up at the WG, discussed and consensus was
reached, there's no need for me to rehash it.

Thanks Brian!