Browsers not interoperable for first argument of window.open when undefined is passed

[bzbarsky]

Consider this testcase:

<iframe name="foo" src="http://example.com"></iframe>
<button type="button" onclick="doIt()">Click me</button>
<script>
  function doIt() {
      window.open(undefined, "foo");
  }
</script>

What should happen? Per spec the IDL is:

  WindowProxy? open(optional USVString url = "about:blank", optional DOMString target = "_blank", optional [TreatNullAs=EmptyString] DOMString features = "");

so this should be equivalent to passing "about:blank" as the first argument. And in Safari it is: it loads about:blank in the subframe. But in Firefox and Chrome it's a no-op, which is what the spec calls for if "" is passed as the first argument.

When passing "" explicitly as first argument, all browsers treat this testcase as a no-op, so that part is interoperable.

So the real question is whether the default value should be "" or "about:blank".

Note that if the open call creates a new window, both "" and "about:blank" have the same behavior, because https://html.spec.whatwg.org/multipage/window-object.html#window-open-steps step 13 starts off urlRecord as "about:blank" and then step 14 actually runs in the "empty string and new window" case but does not change urlRecord in the empty-string case. So this only matters for the not-creating-a-new-window case.

[annevk]

@cdumez what do you think? Probably safer to change Safari and the HTML Standard, right?

[cdumez]

Ok, let's update the spec then. Just please file a bug against WebKit so I can align our behavior.

[annevk]

PR at #4851 and test at web-platform-tests/wpt#18536. Review of both appreciated. WebKit bug at https://bugs.webkit.org/show_bug.cgi?id=200882.