Re: [rtcweb] A different perspective on the video codec MTI discussion

[Ben Strong <[e-mail]>]

I've been talking to developers who are interested in building WebRTC apps
regularly for the past year and have drawn the opposite conclusion about
their preferences.

Before going into what they want and why, let me give you a flavor of who
I'm talking to: These are primarily inbound contacts who are interested in
the vLine WebRTC platform. I think they're pretty representative of the
non-legacy use-cases for WebRTC, but almost certainly under-represent the
existing enterprise market.

* Over half are building apps for healthcare or education. I'd say it's a
pretty even split between those categories, and within each category it's a
pretty even split between startups and more established organizations.
Non-profits make up a substantial fraction of both groups.

* Over half are outside the US, with most of the international interest
coming from developing countries (e.g., India, Brazil, and China).

* Not one of them has been interested in enabling video for customer
service calls (though a number have been interested in voice). I mention
this because it's frequently cited as an important use-case and I expected
it to be an important market for us, but I'm beginning to wonder if
businesses actually want this. I'll admit that those who _are_ interested
are more likely to be talking to Jonathan than to me, so take this with a
grain of salt.

* A sampling of other apps people are building: translation services (both
for foreign languages and sign language), social commerce services, and
social apps with all sorts of interesting new interaction models.

Here's what I've heard from them:

* The only questions I have ever gotten about interop with legacy equipment
are in regard to voice. I do occasionally talk to people with existing
investments in video conferencing systems, but they are mostly interested
in moving away from them (huge sample bias here, of course).

* Very rarely does the subject of codecs come up other than the context of
whether the browsers will all interoperate. The only concerns I ever hear
about VP8 quality are from people who haven't actually used WebRTC in
Chrome. The feedback I get from people who have used it is consistently
that they are surprised by how good the video quality is.

* All of them care about having a good mobile experience and
do occasionally express concerns about the battery life and performance
impact of software codecs.

* _Everyone_ is concerned about cross-browser interoperability. It is far
and away the biggest concern I hear about WebRTC, and I can't remember the
last time it didn't come up in a conversation with a customer.

Finally, here's my take on it all:

>From a power and performance point of view, H.264 is undoubtedly a better
option on devices with a hardware codec for it. Implementations on those
devices should by all means support it and attempt to negotiate it.
Implementations on other platforms where it is feasible to support it,
either because the implementation vendor has a license or the OS includes
it, should by all means support it as well, for the sake of allowing mobile
devices with a hardware codec to use it whenever possible.

I'm happy to grant that H.264 is technically superior to VP8 in every way
(though I think the differences are negligible in real-world use). By all
means implementations that wish to should attempt to negotiate it for that
reason, as well.

However, I fail to see how either one of those points has any bearing on
the MTI decision. The whole point of MTI is guaranteed interoperability
across all WebRTC implementations, which is absolutely critical to WebRTC
adoption. The only criteria should be the best quality codec which can be
supported by all implementations. And because some implementations are open
source, it absolutely must be royalty free.

I've already gone on too long here, but I want to make one final point on
open source. There are two new mobile platforms that have the potential to
drive innovation around mobile apps in general and the mobile web in
particular, and both are open source: FirefoxOS and Ubuntu Mobile. Neither
will be able to be WebRTC compatible unless a royalty-free codec is chosen.

I would argue that FirefoxOS and Ubuntu should be of special concern to the
working group for three reasons: 1) They support web apps as first class
citizens. Since the main goal of WebRTC is to make it possible to build
rich apps entirely with web technologies, I would hope that the innovators
in that arena will be able to support it. 2) I expect that some of the
areas where WebRTC will have the biggest impact (e.g., telemedicine in
rural India) will be most concerned about using a low-cost, open source
platform. 3) Even without every being market-share leaders, open source
platforms tend to drive innovation across the whole market. The best
example of this is how Firefox kickstarted a decade of innovation in
browsers and web apps. I have high hopes that FirefoxOS will do the same.

Adopting a codec that will more firmly entrench legacy providers at the
expense of some of the most innovative new platforms out there seems like a
bad trade-off to me.

Ben

[this post will probably need moderator approval, so apologies if it shows
up after the conversation has moved on]

--
Ben Strong
[e-mail]



On Wed, Mar 13, 2013 at 5:06 PM, Jonathan Rosenberg <[e-mail]>wrote:

> I’d like to put a different perspective on the video MTI discussion.
>
> Much of the discussion has been around video quality, IPR, and
> standardization status. While those are all important factors, to me they
> are secondary to the core question: how does the choice impact uptake of
> the webRTC APIs and protocols by developers who build applications ontop of
> it? In this regard, I would like to suggest that, at this time, adoption of
> VP8 as MTI will slow down adoption of webRTC by turning off developers that
> would otherwise embrace it if H.264 were selected.
>
> The reason is simple. For many application developers, what is interesting
> about webRTC is NOT just browser to browser communications, but connecting
> users in a browser to users elsewhere - on other devices, in other
> networks. After all, the vast majority of web application developers do not
> have the luxury of a massive social graph, or the luxury of their users
> being parked persistently on their website and thus able to receive an
> incoming call. Many websites that have customer support or service needs
> would love to be able to allow their users to have a video call with an
> agent. However, those agents are probably sitting on existing agent systems
> which are not web based, and it’s almost certainly true that they do not
> today support VP8, but rather H.264. Many developers would probably love to
> connect users on the web with users on mobile apps. Most mobile platforms
> today support H264 based hardware acceleration for decode and sometimes
> encode, but not yet VP8. Developers who want to build business
> communications clients will need to connect those users with other users in
> the business, who may be using videophones, PC clients, telepresence units
> or video room systems, the vast majority of which do not support VP8 today,
> but many of which do support H.264.
>
> The reality is – today’s Internet and IP communications systems are built
> on H.264. And unless the developer cares only about living within the
> island of the web browser – a VP8 based solution will simply not meet their
> requirements.
>
> This may not be true down the road. I applaud Google for working hard (and
> spending much money no doubt) to secure an RF license for VP8 from these
> patent holders. I think they should continue pushing and promoting the
> technology because a  free, high quality video codec would be great for the
> Internet. But today, it is not the video codec of the Internet. And, given
> the relatively early days of video, I am sure there will be video codecs
> after VP8. Maybe H.265, maybe the new video codec being developed here at
> the IETF. And once those codecs become more broadly implemented and
> available on the endpoints that matter, we can revise our specs and mandate
> support for them. But this is not about the web of five years from now, its
> about the web today. And if we mandate usage of a codec which is not widely
> implemented in all the endpoints that matter (not just the browser), I fear
> that it will hold developers back from using webRTC and thus prevent us
> from ever having the opportunity to add these video codecs in the future.
>
>
>
> And so – to the supporters of VP8 – I applaud your efforts and thank you
> for them. Please continue. And when the industry is ready, we can make VP8
> MTI in the browser. But we are not there yet.  I ask you to please put the
> needs of the developers ahead of your own, and do not hold back webRTC for
> the benefit of your ideals around an RF and open source video codec. WebRTC
> is too important for that.
>
> I have one more thing to say - speaking now as a developer.
>
> As some of you may know, I recently returned to Cisco as CTO of the cloud
> collaboration group, which is responsible for Webex. Webex was one of the
> first services to do voice and video on the web, using plugins of course.
> For our business, a key requirement is interoperability with other video
> systems in the Cisco portfolio, including our video clients and
> telepresence units. Those are all based on H.264. Consequently, much as I
> would like to avoid the need for a plugin, the benefits of eliminating the
> plugin do not outweigh the drawbacks of having to transcode from VP8 to
> H.264. If IETF selects VP8 as the MTI codec, this will make it dramatically
> more difficult and expensive for us to use webRTC. If H.264 is the MIT
> codec, it will make it much easier for us to use webRTC.
>
>
> Thx,
>
> Jonathan R.
> --
> Jonathan Rosenberg, Ph.D.
> [e-mail]
> http://www.jdrosen.net
>
> _______________________________________________
> rtcweb mailing list
> [e-mail]
> https://www.ietf.org/mailman/listinfo/rtcweb
>
>