Philip Jägenstedt

HTMLMediaElement#HAVE_NOTHING now shows up, and I don't see HAVE_NOTHING anywhere else, so constants on interface objects (constructors) aren't being filtered out, that's good! I wonder if there are any things...

@bzbarsky, the problem was that some things that weren't CONST_CASE were previously excluded, like in https://github.com/GoogleChromeLabs/confluence/issues/259. Including those is the most important thing. But IDL constants also show up both...

> Well, on the interface object and on the prototype Argh, right you are. That just makes the double counting more certain, don't need to find an instance to have...

Yes, that's unfortunately case, and we even enshrined it in the spec: https://html.spec.whatwg.org/multipage/system-state.html#concept-navigator-compatibility-mode @bzbarsky, do you think that https://github.com/GoogleChromeLabs/confluence/issues/219 is the remedy, or something else?

Chrome 53 is missing from that graph, perhaps that is related.

Having a look at the graphs would be nice before adding it to the deployed web site. Maybe they'll be terrible and we should scrap the idea.

The problem is that execCommand and getElementsByName is on HTMLDocument in Gecko, and on Document everywhere else. In other words, part of this mess: https://github.com/whatwg/dom/issues/221

@mdittmer, do you mean the ability to add comments to APIs, or some merging of prototypes in this case?

Yep, looks like Firefox has `Error.prototype.stack` but nobody else. I don't see `stack` in https://tc39.github.io/ecma262/#sec-properties-of-the-error-prototype-object, is this not defined by any spec? @mdittmer, do we have an umbrella bug for...

Blink has `[ NoInterfaceObject ] interface DOMRectList`, which matches https://drafts.fxtf.org/geometry/#DOMRectList I'm pretty sure this changed in the not too distant past.