Spec the disabled content and IDL attribute for HTMLLinkElement (and HTMLStyleElement)?

[domenic]

This has come onto my radar since I need to figure out if we should add [CEReactions] to the disabled, which is currently not specced.

Previously, in https://www.w3.org/Bugs/Public/show_bug.cgi?id=24284, Hixie says:

I've removed the "disabled" attribute on HTMLLinkElement and HTMLStyleElement, since it's CSSOM specific. It should instead be defined on LinkStyle.

As far as I can tell, it does the same as LinkStyle.sheet.disabled, except if LinkStyle.sheet is null, in which case it returns false and ignores setting. (This is based on what the HTML spec used to say.)

According to https://jsbin.com/vecofi/edit?html,output, the situation is a bit more complex than Hixie makes it sound. In particular, in Chrome, Safari, and Edge, the disabled property controls the presence of the disabled content attribute, and the disabled content attribute can be used to disable stylesheets as well.

In Firefox, this is not the case; there the disabled property controls the disabled-ness of the stylesheet, but there is no content attribute behavior at all.

The use counter for the disabled content attribute in Chrome is at https://www.chromestatus.com/metrics/feature/popularity#HTMLLinkElementDisabled and stands at 0.0410%, which is fairly high.

If we want to go with the majority behavior---which we probably should?---then I think this needs to move back into HTML. Perusing the Chrome code, it seems like the behavior is fairly involved, and involves a tri-state DisabledState enum of Unset/EnabledViaScript/Disabled. It also has interactions with alternate stylesheets. See https://code.google.com/p/chromium/codesearch#chromium/src/third_party/WebKit/Source/core/html/HTMLLinkElement.cpp&sq=package:chromium&q=HTMLLinkElement&l=644

[zcorpan]

I don't have a problem with moving it back to HTML or specifying the content attribute (though it only has an effect for stylesheet I suppose).

Regarding the use counter...
https://code.google.com/p/chromium/codesearch#search/&q=HTMLLinkElementDisabled&sq=package:chromium&type=cs
It seems to me it triggers for pages that set the IDL attribute to true because that adds the content attribute. Pages that use the content attribute in markup or setAttribute might be much lower. I didn't find any bug about this in bugzilla.mozilla.org from a quick search.

[foolip]

I added a bunch of use counters in this general area:

• V8SVGStyleElement_Disabled_AttributeGetter
• V8SVGStyleElement_Disabled_AttributeSetter
• HTMLLinkElementDisabled
• V8HTMLLinkElement_Disabled_AttributeGetter
• V8HTMLLinkElement_Disabled_AttributeSetter
• V8HTMLStyleElement_Disabled_AttributeGetter
• V8HTMLStyleElement_Disabled_AttributeSetter

HTMLLinkElementDisabled is for when the disabled content attribute changes, the others are for IDL attribute getters and setters.

[zcorpan]

The difference between V8HTMLLinkElement_Disabled_AttributeSetter and HTMLLinkElementDisabled is something like 0.003 percent units.

Looking in httparchive (472k pages) I see 130 resources (from 128 pages, so ~0.027% of total pages) containing <link disabled>, of which 126 have "stylesheet", 25 have "alternate". See results at https://gist.github.com/zcorpan/05b2a50c5c854d4b2eee62aa0418ca29