Daniel Eran Dilger
Random header image... Refresh for more!

Podcast: Mozilla, Opera & Ogg Theora in HTML 5

 Wp-Content Themes Techbroadcasting Images Radio Logos Nightowlogo

Gene Steinberg of the Tech Night Owl invited me to talk about HTML5 and the future of the web. Listen and subscribe to the Tech Night Owl RSS feed at:

The Tech Night Owl LIVE with Gene Steinberg

Recent episodes:

July 9, 2009, Mozilla and Opera fight to make Ogg Theora the official codec of HTML 5.
June 18, 2009, Microsoft Enthusiasts
May 21, 2009, talking about Windows 7
March 19, 2009, talking about the iPhone 3.0 SDK
February 26, 2009, covering the Apple stockholder meeting
January 8, 2009, covering Apple’s last Macworld Expo

Earlier episodes I’ve participated on:

Oct 16 08
Oct 2 08
July 31 08
June 12 08
May 1 08
Mar 20 08
Jan 31 08
Jan 3 08
Nov 8 07
Sep 20 07
Aug 9 07
Jun 14 07
Apr 26 07
Mar 1 07
Jan 11 07

  • asa@mozilla.org

    Just listened to your podcast and wanted to mention a couple of things.

    First, YouTube’s claim that serving Theora would melt the Web was completely debunked. See http://weblogs.mozillazine.org/asa/archives/2009/06/theora_video_vs.html for some background. Theora+Vorbis in Ogg could replace the complete catalog at YouTube for pretty much the same bandwidth and a better user experience. Theora+Vorbis in Ogg solidly beats the standard def videos (the majority of YouTube served content) at YouTube at similar bitrates and comes very close to the high def content at similar bitrates.

    [Thanks for your comments, but if you actually look critically at that post and its comments, you’ll find that while Theora might be approaching adequacy for the low quality content Google is doing now (Flash video using an H.263 level encoding and extremely basic baseline profile H.264 encoding), it is nowhere near the potential of what H.264 can already do with a decent encoder; H.264 is also improving dramatically and has all sorts of hardware accelerated encoders available for it. That’s important on the creation end and on the playback end. Theora is worthless to mobile/netbook/appliance users that require hardware acceleration, which are increasingly the audience the web will be targeting. – Dan ]

    (and it’s worth noting that Chrome is implementing Theora+Vorbis in Ogg)

    [Well yes, indicating that Google is not prejudiced against Theora for any political or commercial reasons (as one could claim Apple is). Therefore, Google’s claim that Theora is not usable for YouTube is not as dismissible as Ogg proponents like to claim, as you do so above. ]

    Second, Theora is lower CPU usage than h.264 since it doesn’t apply as many tricks to the file both in encoding and decoding. This means that devices that don’t do off-CPU decoding can actually benefit from Theora compared to h.264.

    [What devices would that be, fast PCs without any need for hardware acceleration? The fact that Theora isn’t very sophisticated (being based on technology that was commercially abandoned in 2001) explains why it is simple, but does not imply some benefit, as the only devices that don’t use hardware acceleration are devices that don’t need to use hardware acceleration, making this argument rather specious.]

    Even so, there’s still a lot of headroom for Theora and the team working on that is making great progress. Not all of the good ideas in video compression are patented. Take a look at the differences between the Theora 1.0 codec from less than a year ago and the 1.1 alpha from last month and you’ll see that it’s vastly better. AND they’ve got four or five additional projects that are in the pipeline for improvements without infringing on known patents. Your claims that Theora is stuck is just wrong.

    [I didn’t claim Theora “sucked” (I don’t write for Ars!), I only pointed out that it is based on commercially non-viable technology, is not supported in silicon, and the path ahead for improvement is already paved with technology that is largely patented. The idea of using Free Software is fine in many respects, but it makes no sense to revert to using outdated technology that is a decade old (a century in tech-years) with the hope that ‘progress will be made.’ Surely you are aware that H.264 is progressing on several fronts, supported by lots of companies working together. It is not that expensive to license H.264, it just isn’t completely free. And as I point out, that isn’t a problem for many other GPL products that rely upon grey market software, so why should it be the straw that breaks Mozilla’s back? If Firefox is going to live off the teat of Google’s search ads, perhaps it just needs to step up its adware efforts or find another partner that can pay for its modern media software licensing. ]

    Third, you mention “the silicon problem” and your understanding of this is just plain wrong. There isn’t custom silicon for encoding and decoding h.264. Please read up on how “hardware” decoding actually works before making those kinds of claims. The real situation is that there’s commodity hardware, a handful of popular chips (just like there are a handful of popular CPUs) with various bits of _software_ running on top of them for handling the different codecs. There’s work underway to write that software for Theora on some of the popular existing chips. Apple or anyone else could easily write more of this for the chips they care about. It’s not rocket science and it _does_not_ require hardware fabs. It’s software. Yes, software. Not hardware. Software. Really. Go read about it. It’s being written and more of it could easily be written. The “no silicon” argument is completely bogus and only being used by people who either don’t know better or have some other agenda.

    [Regardless of of how you like to split your hairs, there is no hardware acceleration available for Theora. You can promise that volunteers will donate their time to write low level software that will might enable support for Theora on some chipsets, but until that happens it’s just hot air. Also, as you point out above, Theora isn’t sophisticated enough to really benefit from hardware acceleration. That’s because it was found commercially non-viable nearly a decade ago. ]

    (I’ll leave your characterization of Opera’s business to someone over there, but you seem to be ignoring Opera Mobile which ships on about 7 million phones every quarter — a full Opera browser) and that Opera Mini actually sends real Web content to phones, including video and will send Theora+Vorbis in Ogg before too long. It’s not an image as you characterized it, it’s just “cleaned up” and compressed for easier downloading processing. See also SkyFire and Danger and others doing the same thing.)

    [Look at Opera’s share of actual mobile web traffic and you’ll understand why I devoted short shrift to the company’s products. It’s nice they are interested in open standards, but the company isn’t really relevant apart from offering a limited mobile browser to a lot devices that don’t actually use the web very much. ]

    Next, this isn’t some “free” religious or political issue. Firefox, which has 300 million users and about a quarter of the Web usage cannot included patented codecs. It’s incompatible with our licensing. Your characterization of this as Mozilla at war with commercial interests and your lumping in Mozilla with open source zealots is just plain wrong.

    [Firefox can’t bundle commercial licensing, but you can prompt users to download the necessary playback tools, whether x264, QuickTime, or whatever. Netscape/Mozilla failed at delivering an operating system; the company needs to realize at some point that it has one popular product: a web browser. Not a soup to nuts solution to every users’ problems. ]

    Mozilla is doing what we did with PNG. We’re advocating a patent unencumbered codec for use as a _baseline_ on the Web so that any one who wants to can create and distribute content and software without having to pay a licensing fee. There’s room for other codecs and Mozilla’s never said there wasn’t. We just think that a _baseline_ that everyone can know works across modern browsers would be a good thing.

    [PNG was not only adequate, it was better than GIF. There is no PNG-style correlation with the issues plaguing Theora, not in quality/suitability, not in hardware acceleration, and not in preexisting use of superior alternatives. If Theora were superior to ISO MPEG standards, this would be a very different situation.]

    All modern browsers support JPG and PNG. That’s a fine baseline pair (JPGs for photographic and other highly lossy content and PNG for lossless.) That’s a good baseline that we just achieved this year (finally with IE 8 shipping full PNG support) that might have come to the web a decade sooner if we’d have cared to specify it.

    Mozilla has been pushing on PNG hard for a decade and can take much of the credit for creating the developer pressure on Microsoft and the results are good for the Web. That’s what motivates Mozilla, not some open source zealotry or some desire to avoid patent fees. Making the Web better and easier to participate in is our mission.

    I personally think the Web needs at least two baseline video codecs and several baseline audio codecs. For video, we need something like Theora (or H.264 if it wasn’t for the patent problems) which uses discrete cosine transforms and is good for highly compressed web videos. We also need something like Dirac or another wavelet compression codec that helps us at the very high end and to smooth over the differences between online “broadcast” video. We need something like Vorbis, which is actually superior to MP3 for lossy audio and we need something like Speex or Silk for voice-specific audio. We probably want a lossless audio codec too.

    Just as JPG and PNG are designed for different types of image content, so are the different video and audio codecs. Having a few of them, each for their own specialized content, is actually a good thing.

    [That is fine and good and admirable, but there are also reasons why commercial vendors like Apple do not use Theora/Vobis/FLAC lossless: patents! Do you really think Apple is against free codecs because it wants to pay licensing fees?

    Apple does have some patent royalties related to MPEG, but they do not offset what Apple has to pay. Apple pays, not to repress FOSS, but because its better business to pay into the pool and have access to the best technology than to adopt some volunteer effort technology that almost assuredly will be hit by submarine patents because volunteer developers do not have volunteer patent attorneys. And meanwhile, while Apple is humoring the FOSS community by dicking around with old crap codecs like Theora, Microsoft will waltz in and force WMV/VC-1 as a far superior technology, killing any future for iPods/Macs/iPhones, etc. Right, upholding FOSS politics is worth losing?

    And if Apple were to appease Mozilla and support Theora as the official codec for web video, what would happen if Theora came under GIF-style patent claims (say, from Microsoft, which has all sorts of VC-1 patents)? Mozilla would cut and run and pick up some other free codec, leaving Apple stuck with a bunch of hardware that wasn’t usable and a patent troll to fight in court. If a patent troll hits MPEG, Apple has allies that are far more significant than Mozilla. That is to say, significant at all. ]

    If the W3C specifies any codecs, they must be patent unencumbered. That’s their requirement, and a good one. I think they should specify a handful of baseline codecs so we don’t end up like we did with PNG where web content producers have to wait a decade for a cross-browser baseline that works.

    [Specifying a baseline codec is not necessary, and does not result in vendors implementing it. Specifying codecs that vendors do not actually support is like writing stupid laws that can’t or won’t be policed. It just gives people reason to ignore the law in general and hold it in contempt. Microsoft will be pushed to support PNG, H.264, and HTML 5 faster if those standards are viable and demanded by consumers and producers than if they are only specified in specifications. ]

    Notice that I’m not saying that Gif or H.264 are bad or evil. If people want to use them, so be it. I just think that we shouldn’t write those into any specification (and the W3C won’t) and I think that most of the arguments against the patent-free codecs are wrong, sometimes because of a lack of information or understanding of this space and sometimes intentionally wrong for purposes of spreading FUD.

    [Fortunately for you, the W3C isn’t being pressured into specifying GIF or H.264, because HTML is not supposed to specify file types. Mozilla is pushing for specifying a baseline, which is unfortunate because it is a commercially non-viable, obsolete, and non-mobile-savvy baseline. ]

    So, I encourage you to go look at the progress being made on Theora and I encourage you to go read up on how off-CPU hardware decoding actually works and I encourage you to read the arguments at the WHATWG mailing list in support of baseline codecs and I encourage you to go read about Mozilla’s mission and success in bringing competition back to the browser space and I encourage you to go read about Opera’s products and plans.

    When you’ve educated yourself, I think you’ll see things a bit differently. I hope you’ll then follow-up on that podcast with a post espousing your newfound knowledge.

    [While I’m getting smart enough to approach Mozilla’s feet to question the great FOSS god, perhaps Mozilla can happen upon a business model so you are not out of a job as soon as Google realizes that it doesn’t have to subsidize the ashes of Netscape anymore just to block Microsoft’s encroachment into web search. Because that’s about a year or two out. Keep releasing stuff like FF 3.5 and it will be a year out. – Dan ]