W3C home > Mailing lists > Public > public-html@w3.org > October 2009

Re: typeof document.all

From: Brendan Eich <brendan@mozilla.org>
Date: Tue, 13 Oct 2009 14:52:57 -0700
To: hallvord@opera.com
Message-Id: <45E0CB2C-C7C8-4966-9DDC-724A7462FD3C@mozilla.org>
Cc: Anne van Kesteren <annevk@opera.com>, Maciej Stachowiak <mjs@apple.com>, HTML WG <public-html@w3.org>
On Oct 13, 2009, at 2:16 PM, hallvord@opera.com wrote:

> Siterer Brendan Eich <brendan@mozilla.org>:
>
>>>> How is that not what you called a "weird result"? It's similar  
>>>> to  what we do in assigning undefined, i.e., treating assignment  
>>>> as  "detecting".
>>>
>>> Having the object evaluate to false does not seem so weird to me.
>
> Just chiming in to confirm what Anne said (for proof and reference I  
> attach a somewhat enhanced demo file we used to compare our doc.all  
> cloaking and support to other implementations).

No problem, I believed Anne! No doubt.


> I didn't quite understand the comment about "treating assignment as  
> 'detecting'" because we do no such thing.

The jaron is odd, but you do. A "detecting" use of document.all would be

if (document.all) { IE-only content here }

We found (see the links in https://bugzilla.mozilla.org/show_bug.cgi?id=259935 
#c0 ) that content also did

  d=document;
  this.ie=(d.all);
and so assignment is "detecting" too, and does *not* detect the  
document.all collection.

Opera does the same thing with a different value, false instead of  
undefined, which is why I wrote "It's similar". It's almos the same in  
effect, in this kind of assignment use-case.


>> Let me get this straight: the only cases where Opera evaluates
>> document.all to false are cases in ECMA-262 that apply the ToBoolean
>> internal helper. Otherwise it evaluates to a collection.
>
> While I haven't looked up ToBoolean in 262 now to confirm, this  
> sounds exactly right.
>
> We seem to have gotten away without more magic, compatibility-wise.  
> I looked through your Google code searches.

See the bug cited above and its links (the geocities one is probably  
gone now; the ylib_dom.js one is still live and it does the "d" and  
"this.ie" assignments shown above).

If Opera dodges bullets by other means, great. I was wondering,  
though, if using undefined instead of false as the masquerade value  
would not be even better, since undefined is falsy but also compares  
== itself and null.

Thanks for the note -- my regards to Chris.

/be
Received on Tuesday, 13 October 2009 21:53:49 GMT

This archive was generated by hypermail 2.2.0+W3C-0.50 : Wednesday, 9 May 2012 00:16:50 GMT