Re: RTCSession object

(I don't suppose this was meant to be private so CCing the list again)

On 25.04.14, 12:58, Robin Raymond wrote:
>
> see inline comments
>
> Emil Ivov wrote:
>>
>> A minor comment. Does this map 1:1 to an RTP session? I don't think so
>> but if it doesn't then how about using a different name? RTCDialog?
>> RTCContext? RTCConnection? RTCPeerConnection?
>
> [RR] Yes, RTCContext or something might be a better name.
>
>> Second, less minor, I understand the need for a default session and
>> trying to spare users from having to mess with it, but am a bit
>> worried that one default session per page might not work well. I am
>> thinking about cases where one page establishes connections with
>> different peers. There are many examples of this today.
>
> [RR] No, it's not an issue across different peers, at least for the
> freezing case because the freezing is based on the combination of the
> local and remote foundations being a match and different peers will
> cause different local and remote foundation pair values. Thus freezing
> only happens within the context of a single peer to peer connection and
> never between different peers, i.e. that case not an issue.
>
>> So I am wondering if this shouldn't be a more obvious and mandatory
>> part of the API.
>
> [RR] I hope we don't find a corner case to mandate a context object
> being needed (beyond an auto defaulted one). Right now I don't think we do.

Yes, I wasn't really that worried about the ICE part. During the last 
call however, Justin mentioned a few other reasons where we'd need to 
use such a context. Not sure I remember them all but synchronisation 
might have been one. CNAME generation too. Was there anything else?

Emil


-- 
https://jitsi.org

Received on Friday, 25 April 2014 11:20:11 UTC