[CSSOM] Separate camel_cased_attribute from CSSStyleDeclaration

[litherum]

In #825 (comment), we resolved to:

1. Remove _camel_cased_attribute from CSSStyleDeclaration.
2. Create another interface with _camel_cased_attribute, somehow mix it into CSSStyleRule? The DOM2Style spec was extremely hand-wavy about this worked.

[zcorpan]

Do you mean mix it into CSSStyleDeclaration?

I don't understand what is being requested.

[tabatkins]

Yeah, CSSStyleDeclaration.

Some context:

Currently, if we change CSSFontFaceRule's spec back to matching reality (giving it an attribute CSSStyleDeclaration style;), it'll come with every CSS property as well. This is nasty and dumb, and since Firefox doesn't do this and hasn't had any problems, clearly something we can avoid.

So, probably pull the _camel_cased_attribute into a separate [NoInterfaceObject] interface, and specify (I guess in prose?) that it's mixed into the version of CSSStyleDeclaration used on CSSStyleRule only. We'll do similar for the set of descriptors actually used by CSSFontFaceRule.style.

(Tho, looking at Firefox, it currently uses CSS2Declarations as the type for CSSStyleRule.style; this interface inherits from CSSStyleDeclaration. Presumably we can just copy that, then, and make both instances uses subclasses instead, to avoid doing weird prose shenanigans.)

[zcorpan]

So we have three different versions of CSSStyleDeclaration:

• The one returned by element.style
• The one returned by getComputedStyle(element)
• The one returned by cssStyleRule.style or cssFontFaceRule.style (or are these also different?)

Should each of these have different subclasses? Is that web compatible? What do browsers do?

Also note that it's not sufficient to separate only _camel_cased_attribute, there's also _webkit_cased_attribute, _dashed_attribute, and cssFloat.

[gsnedders]

We resolved this in #5649.