Posts Tagged sdp
I’ve spent many years of my career working on interoperability in communication systems. Back in the dark ages, I did SS7 interoperability testing. During my CLEC days, I ran a test lab that tested optical, telephony, and ATM/Frame Relay equipment. I’ve spent many years working on interoperability issues with SIP, starting with the SIP call flows (RFC 3665 and RFC 3666) and then SDP Offer answer (RFC 4317). I’ve also been to many SIPits (SIP interoperability events run by the SIP Forum), testing voice and video interoperability.
WebRTC poses some interesting interoperability challenges, but I am hopeful we will get it right.
There are four different areas of interoperability: browser, protocol, codec, and offer/answer. Lets go through them one by one.
Browser interoperability is about aWebRTC application or site working the same regardless of which browser the user is using. In the past browser interoperability was just a browser/server issue, but with the peer-to-peer media and data channel flows of WebRTC, this is now a browser/browser issue. The good news is that there are only a handful of browsers, so the interop matrix is not too large. The bad news is that there are signs of discord already in pre-standards implementations. For one thing, all browsers must utilize the same APIs, or else WebRTC will be a major headache for developers. Of course, libraries can hide this complexity from developers, but this will slow down deployment and produces some needlessly bad user experiences. If we see one browser vendor using their own APIs instead of using the standard ones from the W3C, then we will know that someone is playing company games at the expense of the Internet users of the world. Hopefully this won’t happen, but it if does, users will and developers will likely move away from that browser.
Protocol interoperability is a major concern for WebRTC. In the past, browsers didn’t implement many protocols – everything used HTTP (Hyper-Text Transport Protocol). Today, browsers are doing more, including WebSockets, and will soon move to the next version of HTTP, 2.0. With WebRTC, the browser RTC Function has to implement multiple protocols including RTP, ICE, STUN, TURN, SCTP etc. These protocols define “bits on the wire” and “state machines” that ensure that interoperability works. For browser-to-browser media and data channels to work, browsers must implement these protocols and carefully follow the standards. If they don’t the whole industry will suffer. There are some issues today with the pre-standard WebRTC browser implementations. For example, one browser today implements a proprietary STUN client that will not work with standard STUN servers. Browser vendors will need to take protocol interoperability very seriously, and recognize that this is something new for them and that they need to follow industry best practices and approaches.
Codec interoperability is about ensuring that media sessions don’t fail because there is no common codec supported on both ends of the session. There are so many codecs in use, and every vendor and service provider seems to have their own favorite one. Fortunately, we should be able to avoid this for audio codecs. The IETF has recently finalized the Opus audio codec for speech and music, published as RFC 6717 this month. It really is a fantastic codec, much better than all the rest, making it an easy choice as one mandatory to implement (MTI) codec for WebRTC. Opus is also available as open source. The other MTI codec is G.711, also known as PCM, which provides interoperability in the VoIP and telephony world, and is also needed for interworking with the telephone network. Video codec choice is much more difficult. While H.264 is widely used today, there are no open source implementations or royalty-free licensing available for browsers or implementors. As such, it is very difficult to see how it could be chosen as a MTI video codec. Google’s VP8 video codec is proposed as an alternative, and is available in open source. However, there is much uncertainty about the licensing status of VP8. Should WebRTC deploy without common video codecs, this again could result in interoperability delays.
Offer/answer interoperability is perhaps the least understood, but most important area. Offer/answer refers to the negotiation of codecs, parameters, and settings for the media session or data channel between the two browsers. Even if both browsers use common APIs, standard protocols, and common codecs, if they are unable to successfully negotiate and configure their media or data channel, the connection will fail. WebRTC uses Session Description Protocol (SDP) to do this offer/answer exchange. The pre-standard WebRTC implementations are, frankly, a mess in this area. Their SDP is not standard, and not interoperable with anything else. It will take a lot of work to get this right, and we all must insist that browser vendors support standard offer/answer negotiations.
Occasionally, it is suggested that perhaps offer/answer would be easier if we didn’t use SDP. We all know and hate SDP, and it is ugly and awkward to use. However, it has taken over a decade’s work and experience to make it work, and any replacement would likely take that many years to get to work. And, in addition, since much of the standards-based VoIP and video world uses SDP, it would need to map to SDP as well. I can’t see this helping interoperability in any way. Previous efforts to replace SDP failed (anyone remember SDPng?) and I think anyone advocating replacing SDP needs to explain why a new effort wouldn’t meet a similar end, and why this effort wouldn’t take a decade. Also, the complexities of offer/answer relate to the complexities of negotiating an end-to-end session, and the actual syntax of the descriptions are a very small part of the complexity.
So WebRTC definitely has some interoperability challenges ahead of it. Fortunately, there are
many experienced engineers who are participating and helping with the effort. As long as the browser vendors take this seriously and don’t play games, I think WebRTC will have good interoperability, which will benefit web developers and web users alike.
If you are interested in WebRTC, you might like my new book “WebRTC: APIs and RTCWEB Protocols of the HTML5 Real-Time Web” published this month by Digital Codex LLC.