Re: [css3-images] Proposed Gradients changes

On Monday, November 29, 2010, 4:04:40 PM, Alan wrote:

AG> On 25/11/2010 5:13 PM, Chris Lilley wrote:
>> On Thursday, November 25, 2010, 4:56:03 AM, Alan wrote:
AG> [snip]
>> http://www.w3.org/TR/css3-color/#rgb-color


>>>> If the sRGB value (-20,270,250) falls within the device gamult
>>>> it will be displayed, otherwise it will be clipped so that it
>>>> falls inside the device gamut.


>> AG>  So why doesn't CSS3 color say this?

>> It does. Same section:

>> "Values outside the device gamut should be clipped or mapped into the
>> gamut when the gamut is known: the red, green, and blue values must
>> be changed to fall within the range supported by the device"

AG> CSS3 color does not say what is displayed for values outside the device
AG> gamut.

Correct; this is intentional. Hard clipping can be done various ways (hue preserving, lightness preserving, and so on) which may be chosen by a color management system. So the precise method of clipping, and the colour that results, is not specified; only that the out of device gamut colour must be clipped.

AG>  Also the example of what is outside a sRGB gamut in the spec is a
AG> printer.

Yes.

AG>  A printer uses CMYK colours 

Many do, yes. They still have a range of colours which they can produce (the gamut). That gamut can be measured, and visualised in a suitable colourspace like CIELAB, regardless of whether the printer uses CMYK or some other system.

AG> which is the opposite of Adobe RGB
AG> color space which is RGBW.

No, its not 'the opposite'. If you are thinking of the oft-repeated but wildly inaccurate set of equations:

C = 1-R (and so on)

then that is a very poor way to convert from RGB to CMYK.

However, regardless of the method of conversion, a printer has  a device gamut just like a screen has a device gamut. Colours can be inside or outside that gamut. 



>> AG>  I can not answer this myself since there is no CIE colorspace
>> that maps AG>  color to a x, y and z matrix

>> Not sure what you mean by that (and I suspect you mean X Y Z, or
>> possibly x y Y, but not x y z).

AG> I talking about x, y, z matrix which uses Cartesian coordinates in three
AG> dimensions [1] or Geometric representation [2]. 

Thanks, I know what a three dimensional color space is :) my point being that this is CIE X Y Z (note the capitals). They have a different meaning from lowercase x y z, which are normalized forms such that x+y+z=1. It is that equation which allows z to be discarded, producing a two dimensional chromaticity diagram.

AG> CIE 1931 x,y 
AG> chromaticity space is is not sRGB colorspace. 

Well, that is certainly true.

AG> the former is shaped like
AG> a convex horseshoe and the later is a cubed prism with x, y, z coordinates.

So, you mean CIE XYZ as I thought. Going back to your original question then, you assert that there is no CIE colorspace
that maps  color to an X, Y and Z matrix. But this is incorrect, since CIE XYZ is indeed a colourspace and thus, points in that space represent colours. 

AG> What I am asking is if there is actual color at all points between the
AG> rings of monochromatic colors and the line of purples and the top of the
AG> x, y, z prism which is white. Please view what I trying to express in 
AG> mere words.

AG> <http://css-class.com/images/xyz-colorspace.png>

Sorry, but neither your words nor your diagram is especially illuminating. Does the gray cube represent an RGB gamut of some kind? That gamut is no longer a cube when drawn in XYZ space.

AG> If a gradient was premultiplied, then any color within  the x, y, z 
AG> prism goes *directly* from that point of color to the point at the top
AG> of the x, y, z prism. Appearing over a white background, a yellow to 
AG> transparent gradient is mixed like pigments.

I think you are getting pretty mixed up here, if by pigments you mean a subtractive colour space. CIE YZ is an additive space.

AG> To base a spec for gradient on the mixing of colour like pigments is 
AG> like testing gradients of color to transparent just with a white even 
AG> background and being amazed that the midpoint between yellow and 
AG> transparent is a pale olive but them believing that it is gray. This is
AG> not a paint class.

Maybe you should start again with what your actual point was. 


-- 
 Chris Lilley   Technical Director, Interaction Domain                 
 W3C Graphics Activity Lead, Fonts Activity Lead
 Co-Chair, W3C Hypertext CG
 Member, CSS, WebFonts, SVG Working Groups

Received on Monday, 29 November 2010 15:55:44 UTC