(Was the summary supposed to say "*un*expected objects" - or is it just me not understanding WebIDL terminology yet? :-))

I do try to keep the terminology as confusing as possible. :)

Another way to hit the same problem is of course to do 

document.alert=window.alert;
document.alert('If a document pops up an alert and nobody sees it, does it still pop up?');

The other case where this come up is this:

  function f() {
  }
  f.prototype = document;
  var d = new f();
  d.write("What happens here?");

Some UAs throw in this situation; some actually invoke the native write function on the document.  I would prefer throwing, as the represetative of one of the UAs which currently does NOT throw here but would like to change that behavior.

I've added back the check (which I am sure used to be in there) for calling Functions that correspond to IDL operations so that it throws a TypeError if the 'this' object is not an object that implements the relevant interface.  (This check was already present for most of the other Function cases, like getter/setter Functions.)  Whether an object "implements" a given interface is still kinda hand-wavey, but the intention is to throw for cases like comment 4.  I'll try to come up with a rigorous definition of "implements" as part of bug 12320.

Noticed 

apf.setZeroTimeout=(apf.isIE&&apf.isIE<9)||apf.isO3||apf.isAIR?setTimeout:(function(){ . .

in cloud9ide's source. Now, this might simply be breaking browsers they don't bother testing in..but if this is actually tested code, it seems having a pointer to setTimeout on another object will work in some browsers.

(In reply to comment #6)
> Noticed 
> 
> apf.setZeroTimeout=(apf.isIE&&apf.isIE<9)||apf.isO3||apf.isAIR?setTimeout:(function(){
> . .
> 
> in cloud9ide's source. Now, this might simply be breaking browsers they don't
> bother testing in..but if this is actually tested code, it seems having a
> pointer to setTimeout on another object will work in some browsers.

That doesn't appear to work today:
http://people.mozilla.org/~cmccormack/tests/settimeout-this.html

I've reached the same conclusion - site is probably simply broken..