Re: Server-oriented stack (Was: Summary (updated)...)

Il giorno Wed, 29 Jan 2014 15:35:36 +0100
Emil Ivov <emcho@jitsi.org> ha scritto:

> Hey all,
> 
> On Wed, Jan 29, 2014 at 3:25 AM, Justin Uberti <juberti@google.com>
> wrote:
> >
> > Missing server-oriented version of WebRTC
> > Multiparty, recording, broadcast
> 
> I was wondering if we had a plan as to how exactly we were going to
> take on these two (if at all). The reason I am asking is because I
> think we might already have most of the components in place within the
> Jitsi community (FOSS):
> 
> * media stack and srtp support with libjitsi (with DTLS from
> bouncycastle)
> * ICE support with ice4j.org
> * a video router (SFU) with Jitsi Videobridge
> 
> The one part that's currently missing is recording ... but we
> shouldn't be long...
> 
> None of the above components rely on any part of the existing
> webrtc.org stack, which should also address the interoperability
> concerns that otherwise arise with the dominance of  single RI.
> 
> (FWIW, I do agree that working on formally specifying any part of the
> server-side APIs here would be rather out of scope)
> 
> Emil
> 
> --
> https://jitsi.org
> 

I guess that to complete the Java-based picture Emil described,
gstreamer-java could be used to implement recording: although
gstreamer-java only supports gstreamer 0.10 and not the newer, more
stable 1.x, it already implements stuff that may be used for the
purpose (RTP, VP8/Opus, transcoding, file formats, etc.).

Anyway, just FYI I'm working on a WebRTC server implementation myself as
well. Since it's completely written in C, it doesn't make use of any
webrtc.org code either. The idea was to write a general purpose gateway
that only took care of the protocols, and then leave the application
logic to server-side plugins, which is what I've been working on so far
lately. It's not ready yet, but it will be open source (github
probably) which should hopefully help in widening the range of
plugins/applications based on it, including recording, broadcasting and
so on.

That said, I agree with Emil that a server-side API is not really
required. After all, the concept behind WebRTC is that there's no
client and so server, just peers: the logic behind the peer (be it an
application or a person) is really out of scope.

Lorenzo

Received on Wednesday, 29 January 2014 15:33:53 UTC