- From: Adam Bergkvist <adam.bergkvist@ericsson.com>
- Date: Tue, 14 Jan 2014 09:16:56 +0100
- To: "piranna@gmail.com" <piranna@gmail.com>
- CC: public-webrtc <public-webrtc@w3.org>, Luis López Fernández <luis.lopez@urjc.es>
On 2014-01-13 08:48, piranna@gmail.com wrote: >> I'm hesitant to do these changes since they don't enable, or make it >> significantly easier to enable, anything that's not possible with the >> current API. I see these as convenience things that some developers find >> useful; and as you show, it's rather easy to derive these features from the >> information provided by the current API. I would be more concerned if you >> couldn't write a shim for this in 50, or so, lines of code. >> > I absolutly agree that they could be viewed as convenience things > useful for developers (that are the ones that they will use the API at > the last point!), but for example, the list of current DataChannels is > also to make the API homogeneous regarding to the fact that you can > get a list of the current local and remote streams, list that you > could also be able to build externally counting the calls to > addStream() and listening to the addstream event. > There's a difference here because the APIs are different in many ways. For example, you add a MediaStream, created somewhere else, to an RTCPeerConnection. When you don't want it to be associated with this RTCPeerConnection anymore, you remove it. So you're working with a list of associated streams. On the other hand, you create DataCannels with a factory method. A DataChannel can only be associated with the RTCPeerConnection that created it and it can't live on it's own (like a MediaStream). Having a list of DataChannels would change the rules for garbage collecting the object. We would need a corresponding destroy()-method on the RTCPeerConnection or similar. I don't think we want to go there. /Adam
Received on Tuesday, 14 January 2014 08:17:20 UTC