CSSImageValue intrinsicWidth/Height should be unsigned long

[annevk]

It seems like making these doubles is a mistake?

Wouldn't it be better to make these match (also in name) the naturalWidth/Height attributes of the img element? Seems slightly better to have consistency here? Alternatively perhaps just not expose these?

[tabatkins]

Ah yes, we should probably match up.

I note that img sets its properties to 0 if they don't exist, rather than making them nullable. Do you know if there's a particular reason for this? An intrinsic zero width/height is an actual thing that you can have, so it's not a very effective sentinel value to mean "stretchy in this dimension".

Right now I'm inclined to make them unsigned long?.

[tabatkins]

In the interest of giving myself time to get all this right, I've severely trimmed down the API for this level (#716), and this issue is now moot. I'll move it to level 2.

[annevk]

I'm not sure.

[tabatkins]

K. Absent any further convincing reasoning to match HTML exactly, then, I'll make it nullable. CSS has a lot more stretchy images than HTML tends to have, after all (every gradient!).

[annevk]

I wonder if we can still change HTML or if we want to support lack of height/width differently. It's relevant in HTML too with SVG and such.

[AmeliaBR]

Given the discussion in whatwg/html#3510, and the evidence that current browsers are so inconsistent that there is likely no compat issue with changing the HTML behavior, I would argue for harmonizing both the behavior and the names (i.e., "natural" instead of "intrinsic").

To touch on @annevk's initial comment: with pixel density adjustments, it is conceivable that the integral dimensions are not integers when measured in CSS pixels. But I think consistency should probably outweigh mathematical precision in that case.