Re: [css-colors] Specify the System Colors colors

On Fri, Aug 30, 2013 at 8:11 AM, L. David Baron <dbaron@dbaron.org> wrote:
> On Friday 2013-08-30 12:35 +0200, Simon Pieters wrote:
>> [[
>> User agents must support these keywords, but should map them to
>> "default" values, not based on the user's OS settings (for example,
>> mapping all the "background" colors to white and "foreground" colors
>> to black).
>> ]]
>> -- http://tabatkins.github.io/specs/css-color/Overview.html#system-colors
>
> I tend to think this "should" isn't a good idea.
>
> For a start, the system colors are still needed along with
> 'appearance', given the way appearance is actually (interoperably, I
> think) implemented.  (Was the disagreement between the
> likely-interoperable implementations and the spec part of why it got
> dropped from css-ui, post CR?)
>
> I don't think Gecko would be able to honor this should, both because
> we need the correct values in our UA style sheets and because we
> need them in our UI.
>
> (That said, Gecko also has a number of longstanding prefixed
> extensions to address issues such as those raised in
> http://lists.w3.org/Archives/Public/www-archive/2013Aug/0027.html
> The working group removed these additions from the spec, I believe
> because it wanted to deprecate the entire set.)

I share Dirk's concern.  I didn't think the idea that the System
Colors were bad and security-harmful was anything controversial - I
assumed that it was something already known and accepted.

I don't think we should let UA stylesheet concerns dictate what values
we should expose to authors.  For example, Blink just added support
for a -internal- prefix, which is only recognized in the UA stylesheet
(and soon, other "browser-created" pages like the New Tab page).  This
lets us do things with CSS when it's easier than doing without, but
avoids exposing the properties to web content.  If Gecko did something
similar, it could move its UA/UI usages over to the prefixed colors,
which are accurate according to system settings, while still exposing
just the neutered set to the web.

>> Using just black and white seems unnecessarily boring. It's also not
>> so clear which ones are backgrounds and which are foregrounds.
>
> There's a clear separation between foregrounds, backgrounds, and
> borders.

If we don't go with a specific set of colors from zcorpan's testing,
I'd be okay with adding in gray for "border" colors, and categorizing
the set into one of the three sets for ease of use.

~TJ

Received on Friday, 30 August 2013 20:21:56 UTC