Re: [rtcweb] Requesting "SDP or not SDP" debate to be re-opened

[Roman Shpount <[e-mail]>]

One thing I got to mention is that there is already an API used internally
in current WebRTC implementations: Voice and Video Engines. Both of those
take external network transports (which also have their interfaces defined
and abstract ICE/SRTP/DTLS-SRTP) and allow to specify what codecs to be
used to receive/end media and how this media is encoded. The APIs provided
by these components are fairly typical for the media library APIs that are
used to build SIP and other real time communications clients. In fact these
very APIs were used by (at least at some point in the past) Skype, Google
Talk, Yahoo Messenger and number of other real time clients. I think the
best way to build a usable API is to try to wrap this in JavaScript as
closely as possible instead of building something extremely complicated on
top of them. If we do this, we can create something that will make the job
easier for both browser developers (no complex, hard to test, code which
implements O/A) and for application developers (clear API level contract
with clear mechanisms to control things happening with the media). We
already know that these APIs work for what we need. Let's stop trying to
hide them behind O/A complexity, and then try to invent multiple creative
ways to implement exactly the same functions through all the layers we
built on top.

_____________
Roman Shpount