Re: [css3-images] remaining gradient issues

On Jul 26, 2011, at 12:17 PM, Brian Manthos wrote:

> I’ll leave it to Tab to capture the resolutions from Monday
>  
>  
> Regarding corner-to-corner gradient renderings (regardless of syntax)…
> I recommend (in v3 or v4 of images module) changing the mapping of corner gradients as follows:
> (1) the “gradient line” has a slope of w/h or -w/h depending upon which “ordered corner pair” is indicated
> (2) the “gradient segment” is a portion of the gradient line which (a) has a center point in the middle of the box and (b) has length equal to the length of the diagonal of the box
> (3) the “gradient vector” is the defined as the gradient segment with one of the “end points” designated as a start point
> (4) the rendering for corner-to-corner linear gradients is identical to the rendering produced when using the alternate (angle) syntax with the angle embedded in the “gradient vector”
>  
> Note to spec editors:
> Note my use of 2 new terms: “gradient segment” and “gradient vector”.  We should consider revisiting all references to “gradient line” in the specification.  Some of them have length and location (and thus are segments) and some have direction (and thus are vectors).

I understood very little of what you describe above. In any event, are you saying we should have a different rendering in a future version of gradients than in what this version ends up with once it makes it to REC?

> Regarding Brad’s proposal “add ‘from’ keyword”…An equally attractive and unattractive proposal is to add the ‘to’ keyword instead.  I recommend against both approaches.

Why? It is a simple, clarifying change in existing keyword text in a simple, usable model. I don't understand what you have against it. If 'bottom left' was nearly good enough, but a little ambiguous, then what do you have against adding a word to clarify the meaning?

------

The part of my message about prefixes was _just_ about whether we should temporarily drop all or some of the keywords, and I was pointing out the practical point about the impact this could have on authors. I see no reason to ignore that impact on ideological grounds, when it could inform our decision. Now it could be, as Tab said in the conversation, that no implementor would simultaneously change the meaning of degrees and remove other ways to get a horizontal or vertical gradient in their prefixed version. I hope he is right, but I am not yet convinced he is. If he is not, and such an implementor is expected to follow the spec in another version of their prefixed value, then this does inform my view that we should not remove the alternative and reliable way to get a horizontal or vertical gradient. It does not gain us anything to ignore what implementors do in their widely published experiments and how many thousands of authors benefit from that.