The results of Dialogic's survey on WebRTC Codecs are in! We targeted the survey at Dialogic's developer community and received over 60 responses in less than a week. Respondents included an equitable mix of developers from traditional service providers, established systems integrators, and newer upstarts. Thank you to all who participated as your feedback is invaluable and clearly demonstrates that the majority of the WebRTC
community wants mandatory video codecs, favors royalty-free codecs, and
realizes transcoding between disparate codecs will likely be necessary.
The survey questions were:
Please read below for the results and my observations.
One could argue that WebRTC has survived this long without a mandatory requirement and market forces will ultimately dictate which codec should be used. There is no mandatory video codec for today's room-based video conferencing systems, yet the market settled on H.264 for the vast majority of today's equipment.
On the other hand, guaranteed interoperability between WebRTC-based browsers and non-browser applications is critically important for any application that targets a broad audience. Lack of interoperability is certainly an impediment to WebRTC adoption by many. The majority of our respondents (62%) agreed with this sentiment, indicating they think there should be a mandatory video codec.
As discussed in my last post, the battle over WebRTC video codecs has been very contentious with many unexpected twists and turns. Many of Dialogic's existing customers are heavy H.264 users, so I expected to see that in the results. That's not what happened - 74% were in favor of VP8 being mandatory in some form with only 21% favoring H.264-only. It looks like most do not like the implications of a codec that is not intended to be royalty free.
Interestingly, most indicated they preferred to see both VP8 and H.264 mandatory. This approach allows smaller application developers to choose either VP8 or H.264 but guarantee either will work with browsers that implement WebRTC (eventually all of them).
I should note that the IETF is reviewing its options for how to reach consensus for a mandatory codec. They are accepting new proposals for the options to consider until November 27th. Currently all the options above are listed with some additional ones that add or remove the mandatory criteria from some entities, such as "Browsers MUST support both H.264 and VP8, other entities MUST support at least one of H.264 and VP8". See here for more details.
Our survey target list contains a mix of developers who specialize in many different types of applications including both voice and video, so I expected to see a mix of usage.
G.711 has been around since 1972. It is the lowest common denominator codec for voice communications. Given this, it was no surprise it is the most common codec in use today. On the video side, H.264 is the predominant video codec today as mentioned earlier. It follows that H.264 would also be very popular which it was. However, consistent with the previous response, VP8 also scored very highly indicating it is rapidly gaining adoption despite only being a few of years old.
The future is grim for G.711 as it approaches its 42nd birthday - it had the largest drop in planned usage, presumably as developers switch to the more versatile Opus codec or video communications instead of voice. Consistent with previous responses, on the video side there was a slight drop in H.264 usage with a 17% increase in planned usage for VP8 giving it a slight edge over H.264.
In many ways, transcoding is a double-edge sword. From one perspective the need to transcode shows there was a failure to standardize interoperability, leaving costs the industry must pay for. From another perspective, transcoding helps to solve inevitable codec mismatches that occur in real networks that are evolving faster or differently than the consensus-oriented standards bodies. Our survey results show most respondents were fairly pragmatic, falling somewhere in the middle - a little more than half recognized transcoding is needed, but it should be limited.
A significant percentage thought that transcoding was ok for all calls. It is important to remember that this is the norm in many telephony environments.
I was a little surprised, but encouraged to see strong preferences for the newer, royalty-free VP8 codec. Clearly this is an indication that developers are looking to innovate beyond traditional conferencing applications. This is a great thing since existing conferencing services applications are already under threat by newer, more modern options. WebRTC will only accelerate this trend. Respondent developers are seemingly looking to use this to their advantage. As a new Dialogic employee, I look forward to learning more about what they are creating.