For our initial testing on Windows, we would like to enable WebM/VP9 for just one Windows version. This will make diagnosing video problems easier and ensure that we are still testing H.264 MSE on Windows. We've chosen Windows 10 because it is a large minority of Firefox users who probably have modern PCs (and are thus less reliant on H.264 hardware decoding for adequate video performance).

Comment on attachment 8679828 [details] [diff] [review] [MSE] Enable WebM MSE by default on Windows 10. Review of attachment 8679828 [details] [diff] [review]: ----------------------------------------------------------------- Did you attach the wrong patch? This patch is about 4K video, not Windows 10. ::: dom/media/mediasource/MediaSource.cpp @@ +85,5 @@ > IsWebMForced() > { > +#ifdef XP_WIN > + // Disable 4k video on windows vista since it performs poorly. > + if (!IsWin10OrLater()) { Your comment says Vista, but this !IsWin10OrLater() check will also include Windows 7, 8, and 8.1. Which is correct: the code or the comment?

sorry, it's my super quick copy/paste from WMF decoder will fix

Comment on attachment 8679855 [details] [diff] [review] [MSE] Enable WebM MSE by default on Windows 10. Review of attachment 8679855 [details] [diff] [review]: ----------------------------------------------------------------- ::: dom/media/mediasource/MediaSource.cpp @@ +85,5 @@ > IsWebMForced() > { > +#ifdef XP_WIN > + if (!IsWin10OrLater()) { > + return true; The comment is saying > // Returns true if > ... > // 3. Windows 10 or later. but the code is doing the opposite.

Comment on attachment 8679855 [details] [diff] [review] [MSE] Enable WebM MSE by default on Windows 10. Review of attachment 8679855 [details] [diff] [review]: ----------------------------------------------------------------- ::: dom/media/mediasource/MediaSource.cpp @@ +84,5 @@ > static bool > IsWebMForced() > { > +#ifdef XP_WIN > + if (!IsWin10OrLater()) { I think we should we wrap this check in `#if defined(WIN_XP) && !defined(RELEASE_BUILD)` (i.e. only force WebM for Windows 10 in Nightly and Aurora channels) until we are ready to ship WebM by default. But Anthony may have other plans. RELEASE_BUILD is #defined (counterintuitively) for both Beta and Release channel (because they both have have "Firefox" release branding): https://wiki.mozilla.org/Platform/Channel-specific_build_defines

oops :)

(In reply to Chris Peterson [:cpeterson] from comment #6) > Comment on attachment 8679855 [details] [diff] [review] > [MSE] Enable WebM MSE by default on Windows 10. > > Review of attachment 8679855 [details] [diff] [review]: > ----------------------------------------------------------------- > > ::: dom/media/mediasource/MediaSource.cpp > @@ +84,5 @@ > > static bool > > IsWebMForced() > > { > > +#ifdef XP_WIN > > + if (!IsWin10OrLater()) { > > I think we should we wrap this check in `#if defined(WIN_XP) && > !defined(RELEASE_BUILD)` (i.e. only force WebM for Windows 10 in Nightly and > Aurora channels) until we are ready to ship WebM by default. But Anthony may > have other plans. why that ? If the long term plan is to release webm everywhere, we have to start somewhere... Would be disappointing to have webm all this time only to turn it off on release.

(In reply to Jean-Yves Avenard [:jya] from comment #8) > why that ? If the long term plan is to release webm everywhere, we have to > start somewhere... > Would be disappointing to have webm all this time only to turn it off on > release. We don't want to force users with modern computers to use VP9 software decoder when H.264 hardware decoder is available. When websites like Tom's Hardware benchmark Firefox vs Chrome on a new Windows 10 computer, we want them to compare Firefox using H.264 hardware (or VP9 hardware!) vs Chrome using VP9 software. Anthony agrees that we should only force Windows 10 users to use webm in the Nightly and Aurora channels (i.e. #ifndef RELEASE_BUILD). I expect more Nightly users run Windows 10 than XP. Forcing Windows 10 will give us more test coverage of webm so we are confident it will also work well on XP (on the Release channel).

(In reply to Chris Peterson [:cpeterson] from comment #9) > We don't want to force users with modern computers to use VP9 software > decoder when H.264 hardware decoder is available. When websites like Tom's > Hardware benchmark Firefox vs Chrome on a new Windows 10 computer, we want > them to compare Firefox using H.264 hardware (or VP9 hardware!) vs Chrome > using VP9 software. I disagree with this - we aren't forcing a website to pick either VP9 or H.264 - we offer both, and it's up to the website to choose (of course, Youtube will always pick VP9, so in the end the effect is the same). VP9 is a large technical advancement over H.264, and while it may use more CPU on some systems, the reduction in bandwidth is a huge win for those on weaker internet connections, which is why Youtube and many others want to use it. In addition, it is currently the only unencumbered, royalty-free codec that works with MSE. We should always give websites the option to use a royalty-free video codec if they want, and encourage them to do so. Of course, it's reasonable to take baby steps towards enabling this as the #3 top website is being affected. There is additional work being done to help with CPU usage. But I think the end goal should still be "VP9 everywhere".

I agree with Thomas. If we don't offer VP9 then we're forcing people to use H.264. If we offer both, we're giving sites the choice. And for the reasons Thomas indicates, in many cases those people will be happier if the site picks VP9 *even if they have a HW decoder* because it will look better, even if it chews up more CPU or power.

Comment on attachment 8679867 [details] [diff] [review] [MSE] Enable WebM MSE by default on Windows 10. Review of attachment 8679867 [details] [diff] [review]: ----------------------------------------------------------------- Lets enabled MSE based on an estimate of machine performance instead. I don't think Windows 10-ness is a reasonable proxy.

We're not going to enable VP9 based on Windows version.