bug#41544: 26.3; Possible incorrect results from color-distance

[Mattias Engdegård]

7 juni 2020 kl. 16.30 skrev Eli Zaretskii <eliz@gnu.org>:

>  Background colors affect larger
> portions of display, and therefore a bright background is perceived as
> brighter and a dark background as darker, than when the same color is
> used as foreground.

Yes, but we should ask what that means in terms of a predicate to select 
between them. Of all comparisons I've done, the latest color-dark-p predicate 
appears to work fairly well for selecting backgrounds, in the senses that it's 
never bad.

You are quite right that the optimal predicate may be different, and I'm not at 
all opposed to having two separate functions. They may be identical initially, 
but could diverge as we learn more.

Experimentally, it looks like the case of selecting a black/white background is 
more forgiving: the set of 'ambiguous' colours that contrast well with both 
black and white greater than when selecting a black/white text.

If you could come up with a colour that strongly prefers a black background but 
a white foreground, or vice versa, then that would provide us with more 
insight. If not, it might indicate that those colours are rare.