[css-fonts-4] [varfont] Fonts may have their own internal mapping of keyword names to values

[litherum]

Variation fonts may have their own internal mappings of keyword names to values. For example, a font may state that "condensed" may correspond to a "wght" axis value of "78.2".

CSS may want to have an explicit table of mappings for keywords to values for font-weight, font-stretch, and font-style.

Alternatively, the spec might benefit if the font file is consulted instead. Doing this would separate the notion of a keyword from the notion of a value (so, for font fallback, a keyword might be applied as a different value in two fonts used in the same element).

[litherum]

The OpenType spec lists width percentages of reach of the width keywords.

Value	Description	C Definition	% of normal
1	Ultra-condensed	FWIDTH_ULTRA_CONDENSED	50
2	Extra-condensed	FWIDTH_EXTRA_CONDENSED	62.5
3	Condensed	FWIDTH_CONDENSED	75
4	Semi-condensed	FWIDTH_SEMI_CONDENSED	87.5
5	Medium (normal)	FWIDTH_NORMAL	100
6	Semi-expanded	FWIDTH_SEMI_EXPANDED	112.5
7	Expanded	FWIDTH_EXPANDED	125
8	Extra-expanded	FWIDTH_EXTRA_EXPANDED	150
9	Ultra-expanded	FWIDTH_ULTRA_EXPANDED	200

[Crissov]

Using the ITU radio spectrum as an inspiration, this should be extended to super-*, extremely-* and tremendously-* (and very-* would of course be an alias to extra-*). 😜

I wonder, though, why there’re no 70.7% and 141% √2 presets (as well as 84.1%, 119% and so on). Keywords for irrational values are much more useful than those for simple rational values.

As far as I know, Opentype fonts may also provide default and localized names for indexed features such as ss01, but @font-feature-values does not provide means to access them yet. Anyhow, it’s probably a good idea to make css-fonts work similar in both cases.

[litherum]

Dup of #525