W3C home > Mailing lists > Public > public-script-coord@w3.org > April to June 2015

Re: Should properties on the named properties object be readonly?

From: Boris Zbarsky <bzbarsky@mit.edu>
Date: Thu, 02 Apr 2015 15:26:15 -0400
Message-ID: <551D97D7.9030200@mit.edu>
To: "Mark S. Miller" <erights@google.com>
CC: "public-script-coord@w3.org" <public-script-coord@w3.org>
On 4/2/15 3:04 PM, Mark S. Miller wrote:
> If you initialize [[Writable]] to true, allow assignments, and have the
> resulting value be the value assigned

Hold on.

Let's look at my testcase again:

         <div id="x"></div>
         <script>
           x = 5;
           alert(x);
         </script>

When this script runs, its environment is the global environment.  The 
global has an object (the "named properties object"; it's very like a 
proxy) on its prototype chain which has the special 
[[GetOwnProperty]]/[[DefineOwnProperty]] under discussion.  Its 
[[GetOwnProperty]] for "x" returns a descriptor for a value property 
whose value is the <div> element.

The bareword assignment is happening with the global as receiver.  So if 
the "x" property on the named properties object is writable, then 
assignment to x as above will [[DefineOwnProperty]] on the global 
itself, not on the named properties object.  All good so far.

Now let's consider this testcase:

         <div id="x"></div>
         <script>
           x = 5;
           delete x;
           alert(x);
         </script>

This alerts "[object HTMLDivElement]" in all browsers, as expected: 
nothing ever changed the value of the "x" property on the named 
properties object; we just shadowed it, then deleted the shadowing property.

So it's not that we have the "resulting value be the value assigned"; 
we're just shadowing.

> why not have [[DefineOwnProperty]] also succeed for these cases?

So this is really about cases in which someone does assingment directly 
on the named properties object, not on the global, or does an 
Object.defineProperty on the named properties object.

We _could_ allow people to do that.  It requires a lot more complexity 
in spec and implementations in terms of defining the interaction of the 
proxy-like bits of the named properties object and these user-defined 
properties.  It's a lot simpler to just say that this particular proxy 
never allows properties to be added to it, other than the ones its 
handler wants to expose, than to specify/implement how it should behave 
if it were to allow them.

-Boris
Received on Thursday, 2 April 2015 19:26:43 UTC

This archive was generated by hypermail 2.3.1 : Thursday, 2 April 2015 19:26:44 UTC