W3C home > Mailing lists > Public > public-webrtc@w3.org > May 2014

Re: renaming updateIce

From: cowwoc <cowwoc@bbs.darktech.org>
Date: Fri, 02 May 2014 16:47:36 -0400
Message-ID: <53640468.1030106@bbs.darktech.org>
To: Jan-Ivar Bruaroey <jib@mozilla.com>, Adam Bergkvist <adam.bergkvist@ericsson.com>, public-webrtc@w3.org
On 02/05/2014 4:31 PM, Jan-Ivar Bruaroey wrote:
> On 5/2/14 6:40 AM, Adam Bergkvist wrote:
>> pc.updateIce({ "iceTransports": newValue });
>>
>> The above updates single option, 
>
> That is incorrect. It also sets requestIdentity to "ifconfigured":
>
> dictionary RTCConfiguration {
>      sequence<RTCIceServer> iceServers;
>      RTCIceTransports       iceTransports = "all";
>      RTCIdentityOption      requestIdentity = "ifconfigured";
> };
>
> Absence is not feasible or detectable with default values, just like 
> with default values on c++ function-arguments. Like in c++, the 
> function cannot distinguish between "ifconfigured" actually being 
> passed in and nothing being passed in.
>
> Said another way: webidl binding-code literally fills in the default 
> values before handing it to the implementation (or in the case of a 
> return value, back to content).
>
> Therefore, updateIce as described here 
> http://dev.w3.org/2011/webrtc/editor/webrtc.html#methods
>> 2. If the iceTransports member was omitted and the ICE Agent's ICE 
>> transports setting is unset, set the ICE Agent's ICE transports 
>> setting to the iceTransports dictionary member default value.
>
> is not implementable using webidl-bindings, because there's no way to 
> test whether the iceTransports member was omitted, because it has a 
> default value in webidl.
>
> Said another way: Our spec cannot dictate implementation logic around 
> default values, because that behavior is well-defined.
>
> I recommend treating webidl behavior as invariant in all descriptions 
> of implementation and return values. e.g. remove all mention of 
> default values in prose.
>
> .: Jan-Ivar :.
>

Or, use a different dictionary for the initial configuration and 
updating an existing configuration. The former would have defaults. The 
latter would not. Or just use the Builder pattern :)

Gili
Received on Friday, 2 May 2014 20:48:40 UTC

This archive was generated by hypermail 2.3.1 : Monday, 23 October 2017 15:19:40 UTC