r? @SimonSapin or anyone?

Where would you use this? (0, 0) marks the start of the input. Is this desired? Wouldn’t an obviously invalid location like (u32::MAX, u32::MAX) be preferable? Does it need to be the Default trait rather than some constructor?

Yeah, I don't think (0, 0) it's a good default... Why is this needed?

I'm implementing @font-feature-values rule and I need this to be able to derive Default for FontFeatureValuesRule struct. That means empty FontFeatureValuesRule, I think we don't need to set MAX value or other values for this because these values will be overwritten at some point. Also FontFaceRuleData has something similar. This method can also be changed with Default derive after this.
I'm okay to change this with a constructor named zero or something. But that will make deriving Default for FontFeatureValuesRule or FontFaceRuleData not possible.

I don't think empty() should be implemented using Default, conceptually... But no strong opinion I guess. This seems somewhat easy to misuse, that's all.

So I'm fine with whatever @SimonSapin decides.

That empty constructor should take a SourceLocation as a parameter.

Using Default would allow writing shorter code, but I think in this case it is wrong conceptually. Please write the longer code, sorry :) (Like for FontFaceRuleData::empty, the repetitive parts can probably be generated by a macro.)

Oh okay, I just wanted to shorten the code but if you think it can be misused, we can close this.