__proto__ in Annex B

Geoffrey Sneddon gsneddon at opera.com
Thu Sep 29 11:23:08 PDT 2011

On 29/09/11 19:10, Brendan Eich wrote:
> On Sep 29, 2011, at 6:54 PM, Geoffrey Sneddon wrote:
>> On 28/09/11 00:06, Waldemar Horwat wrote:
>>> Should we standardize __proto__ in Annex B?
>>> MarkM + a few others:  Yes
>>> Waldemar, Doug:  No
>> Unless we have a definite plan that no ES.Next impl will support __proto__, then by all means don't standardize it.
> It is a standard (de-facto), we were just talking about making a non-mandatory normative spec for it. That does carry some weight but not much more than the de-facto standard, IMHO.

Well, de-jure standard is what I meant (even a non-mandatory one).

>> However, seeming the conclusion before was that not all use-cases of __proto__ are handled currently, once they are addressed in ES.Next, there will still be some transitional phase. I'd rather specify it in ES.Next, even if we do then agree to drop it in ES.(Next+1).
>> That said about any transitional phase, if it turns out we can make it [[Writable: false]] in ES.Next, then all the better (thereby only providing a transitional phase for those who can already use Object.getPrototypeOf).
> Making the spec break the existing use-cases won't make browsers break those case. We don't get free-lunch rewrites from busy/absent developers who used writable __proto__ some time in the past for content that is still in service. The spec can only do so much.
> It would be better to remove __proto__ once the horses have found the new, healthy vegetables that replace it, and only then (based on surveys, web scans a la the sawzall study Hixie did [1], etc.).

Indeed. I'm not expecting browsers to change behaviour blindingly based 
upon the spec, but from what I can see the number of instances of people 
writing __proto__ breaks almost nothing, so it may be entirely possible 
to make it non-writable.

But, in general, in any case like this, I'd say the spec should follow 
implementations, not the other way around. If browsers are still going 
to support it, we should spec it.

Geoffrey Sneddon — Opera Software

More information about the es-discuss mailing list