Web-Facing Change PSA: Enabling VP9 by default and removing flag

[Tom Finegan]

Primary eng/PM emails

tomfi...@chromium.org, fgal...@chromium.org, renga...@chromium.org, jlu...@chromium.org

Spec

The VP9 bitstream has been finalized, but the spec is not yet available.

Summary

VP9 is the successor to the VP8 video codec. Chromium has supported pre-release versions of the VP9 bitstream behind a flag since M27. https://codereview.chromium.org/16830004/ adds canPlayType() support, removes the flag, and enables VP9 by default.

Motivation

WebM now supports the VP9 codec, and Chromium was the first to support VP8 in WebM. This continues the tradition, and adds a video codec that produces video of higher quality than VP8 at lower bitrates. This makes the web faster, and all of the cat videos will be that much sharper!

Compatibility Risk

Mozilla already supports VP8 in WebM, but I have no knowledge of their plans with regard to VP9. Microsoft support for VP8 requires installation of additional components to enable support in their browser. The same is true for Apple. In light of the aforementioned points I can't comment with any certainty on the subject of compatibility risk.

Will this feature be supported on all five Blink platforms (Windows, Mac, Linux, Chrome OS and Android)?

VP9 will be supported on Windows, Mac, Linux and Chrome OS initially. Android support is in the works.

OWP launch tracking bug?

N/A

Row on feature dashboard?

No.

Requesting approval to ship?

No. It's already in Chromium. This will remove the flag and enable the final version of the bitstream for everyone.

[PhistucK]

See my comment inline.

☆PhistucK

Well, then, this is really an intent to ship, not an intent to implement. Enabling by default is shipping, unless I am missing something.​​

(Perhaps "no" just meant that you are not requesting an approval, though you do intend to ship.)

[Mark Pilgrim]

On Fri, Jun 14, 2013 at 1:14 PM, Tom Finegan <tomfi...@chromium.org> wrote:
> Spec
> The VP9 bitstream has been finalized, but the spec is not yet available.

> ...

> Compatibility Risk
> Mozilla already supports VP8 in WebM, but I have no knowledge of their plans
> with regard to VP9.

Well, I don't claim to speak for Mozilla, but I'm guessing it would
improve the chances of compatible implementations if a spec were
freely available.

[Tom Finegan]

I was a little bit confused by this point because of VP9 "shipping" behind a flag since M27. In other words, from my perspective, the ship sailed a while ago. If in terms of Blink removing the flag means shipping, then: Yes, we intend to ship. Sorry for any confusion!

 

[PhistucK]

Oh, right. So for future reference, there is usually no approval for implementing (landing behind a flag, even in stable builds). An intent to implement is merely a public service announcement sort of thing that would just help detecting any direct opposition up front (usually there is none for implementing).

Exposing to the web by default (shipping) is a whole other beast and in the Blink world, requires approval from at least three leads.

☆PhistucK

[Jochen Eisinger]

Thanks Tom.

Note that this is merely a PSA at this point, since - while web-exposed - VP9 is a non-blink feature.

best

-jochen

[Elliott Sprehn]

Are you saying there's no spec for this? I sure hope we're not shipping something that has no spec.

- E 

[Adam Barth]

Hi Tom,

I didn't see an answer to Elliott's question below.  (Maybe it got eaten by my spam filter?)

As a general principle, we would like there to be public specifications for features that we expose to the web.  VP8 has an extensive spec [1].  I presume you plan to write something similar for VP9.  Is there a draft we can share publicly?  Can we commit to publishing a draft in a certain time frame?

Thanks,

Adam

[1] http://tools.ietf.org/html/rfc6386

[Paweł Hajdan, Jr.]

Tom, has libvpx made an official release which supports this new format?

Paweł

[Tom Finegan]

On Tue, Jun 18, 2013 at 10:38 AM, Paweł Hajdan, Jr. <phajd...@chromium.org> wrote:

Tom, has libvpx made an official release which supports this new format?

I'm going to go ahead and assume you're asking for a tarball and/or a tag in libvpx.git: No, there has not been an "official" release. 

Also, to answer one of the inevitable follow up questions: The topic has not been discussed by the libvpx team, so I don't have an eta. :)

Tom

-- btw: not ignoring the spec questions, I just don't have an answer yet!

[William Chan (陈智昌)]

I also would like to see a spec first before we ship this to all users by default. And is someone going to add this to the chromestatus.com feature dashboard (I really think we should)? Is there a graceful fallback for browsers that don't support VP9?

[PhistucK]

<video> has <source> and the browser selects the first it supports automatically, so there is a built in fallback mechanism if authors provide the right sources.

☆PhistucK

[Elliott Sprehn]

+1. I'm not comfortable shipping something that doesn't have a spec, doesn't seem like other browser vendors/library authors could correctly evaluate the feasibility of shipping the feature without one.

[PhistucK]

So while we are not comfortable shipping something that has no specification, it seems like we are doing so anyway, in Chrome 29 -

http://blog.chromium.org/2013/07/chrome-29-beta-web-audio-and-webrtc-in.html (see the bottom part of the post)

Public service announcements are nice. No one has to approve.

What gives?

☆PhistucK

[Addy Osmani]

As PhistucK pointed out, V9 appears to have landed despite there being no public specification and a general sense of this being necessary for features we expose on the platform. Has something changed? :/

Is there at least a draft spec available or a time frame for it being made available? Can we get it listed on http://www.chromestatus.com/features?

On Friday, 19 July 2013 07:58:40 UTC+1, PhistucK wrote:

So while we are not comfortable shipping something that has no specification, it seems like we are doing so anyway, in Chrome 29 -

http://blog.chromium.org/2013/07/chrome-29-beta-web-audio-and-webrtc-in.html (see the bottom part of the post)

Public service announcements are nice. No one has to approve.

What gives?

☆PhistucK

On Thu, Jun 20, 2013 at 10:41 AM, Elliott Sprehn <esp...@google.com> wrote:

+1. I'm not comfortable shipping something that doesn't have a spec, doesn't seem like other browser vendors/library authors could correctly evaluate the feasibility of shipping the feature without one.

On Wednesday, June 19, 2013, William Chan (陈智昌) wrote:

I also would like to see a spec first before we ship this to all users by default. And is someone going to add this to the chromestatus.com feature dashboard (I really think we should)? Is there a graceful fallback for browsers that don't support VP9?

On Tue, Jun 18, 2013 at 11:08 AM, Tom Finegan <tomfi...@chromium.org> wrote:

On Tue, Jun 18, 2013 at 10:38 AM, Paweł Hajdan, Jr. <phajd...@chromium.org> wrote:

Tom, has libvpx made an official release which supports this new format?

I'm going to go ahead and assume you're asking for a tarball and/or a tag in libvpx.git: No, there has not been an "official" release. 

Also, to answer one of the inevitable follow up questions: The topic has not been discussed by the libvpx team, so I don't have an eta. :)

Tom

-- btw: not ignoring the spec questions, I just don't have an answer yet!

 

Paweł

On Fri, Jun 14, 2013 at 10:14 AM, Tom Finegan <tomfi...@chromium.org> wrote:

Primary eng/PM emails

tomfi...@chromium.org, fgallig...@chromium.org, renganathan@chromium.org, jlu...@chromium.org

[Max Heinritz]

It's a fair observation, PhistucK and Addy.

To be clear: the Blink guidelines are for the rendering engine, not Chromium as a whole.  We're actively trying to figure out how to bring the spirit of Blink's guidelines to the rest of Chromium in a way that accounts for variation in feature maturity and standards norms. Stay tuned.

In this specific case, I added a row to chromestatus.com.

--
You received this message because you are subscribed to the Google Groups "blink-dev" group.

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/95954a4e-ea0b-4645-a071-5bafe239c533%40chromium.org.
 
 

[Addy Osmani]

Thanks for adding it to chromestatus.com, Max! 

[Mark Pilgrim]

On Fri, Jul 19, 2013 at 11:07 AM, Max Heinritz <m...@chromium.org> wrote:
>> On Fri, Jul 19, 2013 at 3:09 PM, Addy Osmani <ad...@google.com> wrote:
>> Is there at least a draft spec available or a time frame for it being made
>> available?
>

> We're actively trying to figure out how to bring the spirit of
> Blink's guidelines to the rest of Chromium in a way that accounts for
> variation in feature maturity and standards norms.

So... no.

-Mark Pilgrim

[Matthew Dempsky]

Any update here?

I note that the chromestatus.com entry for VP9 links to an expired IETF draft.  According to this webm mailing list post, the bitstream has been finalized, but all of my searching for "vp9 bitstream" just takes me to posts making reference to it being defined.  I haven't had any luck finding the bitstream definition itself.

--
You received this message because you are subscribed to the Google Groups "blink-dev" group.

To view this discussion on the web visit https://groups.google.com/a/chromium.org/d/msgid/blink-dev/CAMNytJB_YvHJMS8nM6LiT8MF4UsDFqDA6KFAH4pP_n1%2Bd1nGsg%40mail.gmail.com.

[Matthew Dempsky]

Ping?  Any news on a VP9 bitstream spec?

[William Chan (陈智昌)]

I'm disappointed there hasn't been any update here. Can someone respond with an update for the VP9 bitstream spec?