Object.prototype.__proto__

[Allen Wirfs-Brock]

On Sep 28, 2012, at 8:00 AM, Axel Rauschmayer wrote:

>>> It is my understanding of B.3.1 that Object.prototype.__proto__’s only use is as a flag to switch off __proto__ for all objects (I’m assuming in settings similar to Caja). Wouldn’t it be better to introduce a real boolean flag, along with a unique name object?
>> 
>> No.
>> 
>> First, __proto__ is what it is, a de-facto standard (albeit with non-configurability in some implementations, a bug to fix), and adding some new switch elsewhere won't change things. So what you propose is redundant (given configurability).
>> 
>> Second, what's wrong with |delete Object.prototype.__proto__| as the explicit, symbol-free (symbol = unique name) way of turning off __proto__ for a given global object and object graph generated within that scope? It's simpler and more direct than setting some other property for a side-effect.
> 
> OK, so this way of switching __proto__ off for all objects is already de-facto standard? Didn’t know, thought it was introduced with ES6. If so, then the de-facto standard obviously beats other ideas. If not then I don’t hate the delete, I just think that overloading __proto__ to be an off-switch in a single object feels slightly wrong.
> 
> Is switching off __proto__ always an all-or-nothing proposition? Or will it be possible to selectively switch it off (e.g. by creating an object whose [[Prototype]] is null)?
> 
>> The only issue remaining with __proto__ is whether it is a magic data property, or an accessor with poisoned get and set pills on reflection.
> 
> Ah, so the spec isn’t (possibly) yet final here(?) At the moment it’s a magic data property (spec-wise), right?


the current draft spec. for __proto__ is at best tentative and will be changed.  It has no significance other than signaling an intent to fully specify __proto__.

Allen