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

> From: Brad Kemper [mailto:brad.kemper@gmail.com]
> Sent: Tuesday, January 26, 2010 6:48 AM
> To: Sylvain Galineau
> Cc: fantasai; www-style@w3.org; Brian Manthos
> Subject: Re: [css3-background] border-radius color transitions using
> gradients recommended but undefined
> 
> Are you getting feedback from authors to the contrary? From my own
> author point of view, I wasn't that concerned about it. I figured the
> UA would either create a reasonable blending of colors at the curve,
> or not blend at all. Is this naiive?

Whatever the UA chooses to do, the author won't have a choice and 
be able to opt out of that difference. As currently specified, authors 
may and likely will get a different rendering - and no blending vs.blending 
is pretty different - whether they like it or not. As this is a problem
they will not run into until they change border-radius from its initial value,
it could be quite unwelcome to those who run into it. 

> In terms of interop issues, I would rate this concern below that of
> the inconsistencies of dotted border styles, where some UAs have true
> dots (IE, in a happy surprise) and others have square dashes instead,
> and no rendering engine is consistent with another in terms of how the
> dots meet in a corner _or_ a curve (at least not the last time I
> checked).

Yes, there are indeed visible issues in this area as well (and much ugliness). 
I would agree it is likely to be a more common problem and see no issue with
prioritizing this and solving it first.

If the spec leaves undefined interop areas that we already know matter to authors then 
we must attempt to fix them.

Some features are so fundamental that authors absolutely *hate* running into interop issues 
when using them. CSS borders is one of those areas (everyone knows IE doesn't do border-radius
for a reason); they should be as dependable as we can make them. Leaving known problematic parts
of it undefined, or implicitly asserting that something is important enough to be recommended but
not important enough to be defined is unhelpful to everyone.

Received on Tuesday, 26 January 2010 17:09:21 UTC