The IDL in the spec says 

  void remove(long index);

but it needs to be

  void remove((HTMLElement or long) child);

and of course the prose needs to be updated too.

Tested in Chrome, Firefox and IE. They all pass the following:

  test('remove', function() {
    var select = document.createElement('select');

    var a = document.createElement('option');
    a.text = 'a';
    select.appendChild(a);

    var b = document.createElement('option');
    b.text = 'b';
    select.appendChild(b);

    var c = document.createElement('option');
    c.text = 'c';
    select.appendChild(c);

    select.remove(a);
    assert.equal(select.firstChild, b);
    assert.equal(select.lastChild, c);

    select.remove(1);
    assert.equal(select.firstChild, b);
    assert.equal(select.lastChild, b);
  });

Does this apply to HTMLOptionsCollection also?

Basically remove() is like add()'s second argument in both cases?

(In reply to Ian 'Hixie' Hickson from comment #2)
> Does this apply to HTMLOptionsCollection also?

Blink/WebKit allows an element in HTMLOptionsCollection remove
Gecko and Trident do not and they only accept an index

I guess I'll go with consistency and hope that Firefox/IE follow suit.

(In reply to Ian 'Hixie' Hickson from comment #4)
> I guess I'll go with consistency and hope that Firefox/IE follow suit.

Makes sense to me

Actually I can't get select.remove() to work with an element. All objects seem to be treated as zero, in all browsers I tested.

   http://software.hixie.ch/utilities/js/live-dom-viewer/saved/2826
   http://software.hixie.ch/utilities/js/live-dom-viewer/saved/2828
   http://software.hixie.ch/utilities/js/live-dom-viewer/saved/2825
   http://software.hixie.ch/utilities/js/live-dom-viewer/saved/2827

It works in your test because you try to remove the first entry, which is treated like zero.

I've updated the spec to convert object arguments to zero.

Please reopen the bug if I made a mistake here.

test from comment 1, for posterity:
   http://software.hixie.ch/utilities/js/live-dom-viewer/saved/2829

Checked in as WHATWG revision r8481.
Check-in comment: Turns out select.remove() and select.options.remove() convert object arguments to zero.
http://html5.org/tools/web-apps-tracker?from=8480&to=8481

You are right. Sorry for the misleading test.

Uh, why did this need a spec change?  IDL will coerce an object to the 0 value of "long" unless it has a valueOf that says otherwise.  That's because ToNumber() will return NaN, and converting NaN to a long gives 0...

In particular, I expect the new spec behavior is not compatible with UA behavior if someone does:

  remove({ valueOf: function() { return 1; } });

BZ is right. I'm sorry for causing all this mess. The old spec was correct and WebKit/Blink behavior was following the spec except that our IDL was misleading.

Lets revert back to the old behavior because that is what Blink, Gecko, Trident does.

Ah, yeah, I misunderstood the ToPrimitive logic in the JS spec. Thanks bz!

Checked in as WHATWG revision r8497.
Check-in comment: Revert r8481 since I misunderstood how ToNumber works on objects.
http://html5.org/tools/web-apps-tracker?from=8496&to=8497