[css-text-decor] Default color of text-shadow

[jfkthame]

https://drafts.csswg.org/css-text-decor-3/#text-shadow-property says that:

If the color of the shadow is not specified, the shadow has the resulting color of the ink that it shadows.

This seems to imply, for instance, that a colored emoji glyph would by default (i.e. no explicit shadow color is specified) cast a multi-colored shadow, as its ink is multi-colored. But ISTM that a "shadow" is inherently monochrome, regardless of the color(s) of the thing being shadowed.

Perhaps this text was written with text decorations in mind, given that text-decoration-color can give different colors to underline, overline, etc.; so the "ink" of a decoration may have a different color than the "ink" of the associated text, and the intent may have been to allow different-colored decorations to cast corresponding different-colored shadows.

But this still seems wrong to me. In the case of text with a colored underline (perhaps even multiple, differently-colored decoration lines), I think the natural interpretation of text-shadow is that it causes the complete (decorated) text rendering to cast a single, monochromatic shadow, and not that it causes the (bare) text to cast a shadow that then has various-colored decorations applied to it.

Also, compare box-shadow, which says:

If the color is absent, the used color is taken from the ‘color’ property.

So a multi-colored <img>, for example, does not magically cast a multi-colored shadow. Neither should multi-colored decorated text, IMO, cast a multi-colored shadow.

Therefore, I suggest the text-shadow spec should be changed to be like box-shadow in this respect.

[upsuper]

cc myself

[kuoe0]

CC myself.

[FremyCompany]

https://jsfiddle.net/bqzwj1ke/ (Edge, Firefox, Chrome)

[FremyCompany]

https://jsfiddle.net/bqzwj1ke/1/

[upsuper]

I guess it means we do want to change the spec to make the whole shadow use currentcolor by default rather than the "color of the ink".

[kojiishi]

So the proposals from @jfkthame and @upsuper are the same, correct? Sounds reasonable to me IIUC.

@FremyCompany's test is interesting...for shadows of decorations, should we honor the used color of color property, or propagated color? Anyone have any preferences?

[upsuper]

The motivation of this proposal was that, having different color in a blur shadow is hard to implement (takes time but doesn't seem to be useful), so I think the proposal is to make the shadow of decoration lines always use the same color as the text.

More formally, the proposal should be something like, when color is omitted, it is treated as if currentcolor is specified.

[FremyCompany]

Then there is also the question of chrome repainting for each text decoration instead of one for the complete geometry. We don't have this issue in Edge as a result of how we do underlining (there can be only one, but its color will vary depending on closest text-decoration:underline parent).

[upsuper]

If we treat omitted <color> as currentcolor, then I don't think there is ambiguity anymore. Everything from the same text-shadow would have the same color.

[kojiishi]

@FremyCompany I'm failing to understand... Can you explain a bit more?

[FremyCompany]

@kojiishi Look at the underline under "bcd" in Chrome in case 1 vs case 2. You will see that the underline in case 1 is "darker" because it seems to be drawn twice (once for the parent "a" underline, and once for the child "span" underline).

[kojiishi]

@FremyCompany oh I see, thank you. Yes, we draw all of nested decorations in order to support different style/position etc., ex. http://jsbin.com/kikuju

[FremyCompany]

@kojiishi I see. But that means that you render using full-color blur, since the underlines can overlap and compose to darker colors. In Edge we try to use the geometry not redraw the text (the same thing we do for background-clip:text, actually).

Also, here is a test case that shows how non-interoperable we are:

https://jsfiddle.net/k53Ld4g3/ (Edge, Firefox, Chrome)

As I understand, this thread seems to be about standardizing Edge's behavior, right?