[css-images] Require dithering in gradients

[muster-mark]

Since this bug was fixed (#4793) dithering has been allowed in gradients.

However in many cases, gradients are essentially unusable without dithering (if you consider their purpose to create a pleasing aesthetic) since it can cause unsightly banding which does precisely the opposite. This is a huge issue in Firefox, which does not apply dithering despite a bug being open for 11 years.

Is there any reason why the specification shouldn't/couldn't require dithering. Care would be needed to define this appropriately, since gradient functions can be used which don't actually create a gradient, and dithering would be undesirable there.

At the very least, could the spec indicate that dithering is a SHOULD rather than a MAY. e.g. something along the lines of "User agents SHOULD employ dithering where necessary to avoid banding effects."

[smfr]

Gradient rendering seems like a "quality of implementation" issue. I don't think the spec should have an opinion on dithering (and doing so may be hard for implementations that rely on underlying graphics libraries to render gradients).

[tabatkins]

Yup, 100% what @smfr says. This is entirely a quality-of-implementation issue. Gradient rendering is usually deferred to the underlying graphics library, so putting a strict requirement there would make impl on some systems unnecessarily difficult, when non-dithered gradients are still okay.

(Also, testing for allowed dithering is already tricky; testing for required dithering seems impossible except in extreme cases.)

[muster-mark]

OK fair enough then, will close this.