RE: [css3-background] border-radius color transitions using gradients recommended but undefined

> From: www-style-request@w3.org [mailto:www-style-request@w3.org] On
> Behalf Of Daniel Glazman
> Sent: Tuesday, January 26, 2010 11:59 PM
> To: www-style@w3.org
> Subject: Re: [css3-background] border-radius color transitions using
> gradients recommended but undefined


> Let's continue the discussion at the technical level, to see if we can
> find a preferred and desired behaviour that is interoperably
> implementable. Speeding up things to avoid another stage back at
> WD level does not seem to me a reasonable argument here.
> If the investigation shows that no interoperable algo can be found
> at this time, then the "undefined _for the time being_ because of
> platform constraints" solution will probably be the best. But since
> this investigation did not really happen, we're unable to resolve now.
> The CR period ends in June and that should be enough to untie the knot.

To be clear, I have no issues with this feature being undefined. I do have a 
problem with it being both recommended and undefined given: 
1. Its impact on the testability of a property that is already popular in its vendor-prefixed form
2. Its uncertain interop impact for authors who do want corner gradients and will fine-tune them in 
one browser only to find this other one does something different enough to be a problem for them
3. Browser vendors being encouraged to 'experiment' through what is now a CR property - no more 
vendor prefix in two implementations I know of - in a way authors cannot opt out of. 

And that's before we even consider issues of technical feasibility which, in my opinion, should be 
investigated *before* we recommend that vendors do something.

I am perfectly fine with this remaining undefined and left for a future version 
of the module. I am also ready, willing and able to experiment behind our vendor 
prefix. But I would indeed prefer to implement this in border-radius. 

Thus keeping corner gradients both recommended and undefined is not and will not be 
acceptable to me until, at a minimum, testability, interop and authors' inability to 
opt out of this unspecified experimental behavior are addressed in some way by the spec.

So while I will apologize for my tone if needed, I do not feel a need to apologize for 
asking too much or being unreasonable.

Received on Wednesday, 27 January 2010 22:02:46 UTC