Xiph, Ogg and Crossing the Chasm

When comparing Ogg Theora to H.264, most reviewers have focused solely on video quality. But that's only a small component of what it will take for Ogg to achieve mainstream success.

I just finished my first foray into encoding Ogg Theora and I have to say that I was impressed by the technology; especially what's coming. But as someone who once marketed technology products for a living, I was very disappointed with the product offering as a whole. Let me explain.

The review process wasn't exactly a painless birth. Here's how it went. I found an excellent book, Dive Into HTML5 with an online component that identified two techniques for producing Ogg Theora files (ffmpeg2theora, which I tried on Windows and www.firefogg.org). Neither produced a file that met my target data rate and/or played smoothly. Then I went to xiph.org (Xiph is the developer of Ogg), clicked an internal link to another Xiph page that led me to "An implementation of our codecs for Apple's Quicktime and Core Audio." I clicked the link and downloaded the XiphQT encoder, which worked within QuickTime Pro and produced the necessary files.

Encoded file in hand, I proceeded with my quality comparisons with H.264, posted the results, and then quickly learned via comment that the product I had used was out of date, despite being recommended on the Xiph site. So, I tried yet a fourth alternative (ffmpeg2theora on the Mac) which worked, though it was a command line process that was cumbersome at best. Then I re-publshed with the updated findings.

At the same time, another author, KeyJ, produced another Ogg related comparison, and also heard from the Xiph folks that he didn't use the most current Theora encoder, despite using the most recent released version. Instead, he learned "Theora development isn’t on Theora trunk, never has been except for release candidates. Current development is theora-ptalarbvorm." If you click over to the referenced site, you'll see a number of files which I guess you could download and somehow work with, but no finished applications that I could find.

All this reminded me of me of my technology marketing days and the difference between a great new technology and a complete product. Every one loves a great new technology, but smart buyers only want to work with a complete product. While this sounds simplistic, understanding the difference will have a profound impact on Ogg's ultimate success (or lack thereof ) and should shape everyone's thinking about the technology as it exists today.

Let's see why in a brief, but very critical, marketing lesson centered around the technology adoption curve and the theory of crossing the chasm. It's old hat to most technology marketing folks, but perhaps a useful review for many others.

The technology adaption curve

New technologies are adopted by various users according to an "adoption curve" first created by Everett Rogers in Diffusions of Innovations and later perfected (some would say) in Geoffrey Moore's foundational marketing book Crossing the Chasm. I grabbed the figure below from an excellent article in Dr. Dobbs by Scott W. Amber.

Briefly, as the figure shows, innovators are willing to take a risk on new technologies, as are some early adapters. Once you get to the mainstream marketplace, however, they need much more substantive proof to adopt. Right in the middle of the Early Adopter group is Moore's chasm, the gap all technologies have to cross to ultimately succeed in a big way. Here's what Moore had to say about the chasm:

"[W]henever truly innovative high-tech products are first brought to market, they will enjoy a warm welcome in an early market made up of technology enthusiasts and visionaries but then will fall into a chasm, during which sales will falter and often plummet. If the products can successfully cross this chasm, they will gain acceptance within a mainstream market dominated by pragmatists and conservatives."

Moore's point is elaborated on in this excellent summary from Norway's Alper Celik.

"The key to crossing the chasm was derived by studying the fundamental differences between the last early adopter and the first pragmatist. While the early adopter would purchase a product that could deliver an 80% solution (seeing it as only 20% more to go), the pragmatist takes the position of buying when it is 100% complete (a ‘whole product’ as Moore puts it) and can be referenced as working within their industry."

Ogg is a promising technology that enjoys an ardent band of vocal supporters, but it will never cross the chasm until the product offering is whole. What would a whole product look like in the codec space? Let's take a look using H.264 as a comparative example.

1.    A whole product is available as an option in all relevant encoding tools on all platforms.

I can produce H.264 in virtually any encoding tool on the planet, and even more important, inexperienced video folks can do the same. You can't do that with Ogg.

Even more important, because Apple is on the dreaded MPEG-LA H.264 patent pool, there's almost an adversarial relationship between some Ogg aficionados and Apple. Here are quotes pulled from some of the comments to my article.

"Please do not use Quicktime [to encode to Theora], as Apple are a member of MPEG LA, so they have a self-interest in promoting h264, and are actively against Theora.   .... As far as I know, there is no available software for Quicktime that implements an even remotely capable version of Theora. If you are at all truly interested in getting a fair and objective contemporary comparison of Theora vs H264, you probably shouldn't be using any Apple platform at all."

Granted, this was only one person, but if it's at all representative of the whole, and Xiph thinks they can succeed without high quality Ogg encoding on the Mac, they're sadly mistaken. In addition, if the software is really that awful, they should pull it from their website.

2.    A whole product plays on all relevant target platforms and devices with all relevant browsers.

Ogg's strategy centers around HTML5, which hasn't been adopted by Internet Explorer, which owns a 60+% market share but can't play Ogg files natively without a plug-in from Google. Why would I want to load a Google plug-in to play Ogg, when I can just play H.264 in Flash? Safari is HTML5 compatible, but can't play Ogg files - only Firefox, Opera and Chrome can. 

Through the Flash plug-in, H.264 plays on (sing along with me, you know the words) 98% of all connected computers. If you don't like Flash, there's Silverlight, and if that doesn't work for you, QuickTime. And, both Safari and Chrome play H.264 natively in HTML5.

H.264 also plays on virtually all portable devices, including all cell phones that support Flash and (I think) Windows Mobile, as well as devices from Apple, Microsoft, Creative Labs and many others. Ogg should play on any device that can load a Firefox browser, but that's about it.

3.    A whole product is supported by extensive hardware for live encoding and accelerated decoding.

Today, I can buy H.264 live encoders from any number of companies -- type in "real time H.264 encoder" in Google, and you get 105,000 hits. Type in "real time Ogg encoder" or "real time Theora encoder" and you get no direct hits. Devices may be coming, but they're not here yet, which means live webcasting with Ogg isn't here yet.

With Flash 10.1, Adobe will utilize graphics hardware to decode H.264 on Windows computers, Android phones and many others platforms, though sadly not the Mac. Apple with Safari is already decoding H.264 directly on supported graphics cards on the Mac. To the best of my knowledge, there are no graphics cards that support Ogg playback acceleration.

4.    A whole product is supported by advanced features like adaptive streaming.

2010 is the year that adaptive streaming will become mainstream, enabling higher quality of service to a broad range of viewers. H.264 is supported in every adaptive streaming scheme, and even has its own scheme called scalable video coding.

Adaptive streaming is in discussions for HTML5, but that's it, and any technology dependent upon HTML5 may be a long time coming. Long story short, if you want to implement adaptive streaming to a broad group of viewers, you can with H.264, and can't with Ogg.

5.    A whole product is supported by the most relevant streaming services.

Another significant trend in 2010 will be the rise of the Online Video Platform, which makes sense for most non-media companies. Why buy server hardware and software, and hire the necessary talent, when you can just pay some third party $1,000/month to do all that for you?

Today, all OVPs distribute Flash only video, and virtually all support H.264. Even Microsoft has issues here, since none (AFAIK) support Silverlight, and none (AFAIK) support Ogg. So even if you want to use Ogg, if you hire an OVP to handle your streaming, it's likely not a supported output.

What's the point?

Let's start with what isn't the point. The point IS NOT that Xiph is doing anything wrong -- all technology organizations have to make the tough transition from technology to whole product, and it's never pretty early on. It's NOT that Xiph or Ogg should be criticized for where they are in the product life cycle, any more than you'd criticize an 6-month old baby for not walking and talking. It's NOT that what they have to accomplish to gain the necessary market penetration is easy, or that there won't be hurdles.

The point IS that Ogg is NOT a "whole product" today and WON'T cross the chasm between early innovators and the mass market until it is. To a degree, the loud cheering that you hear with each Ogg release is at once essential and irrelevant - achieving relative parity in terms of quality is something, but it's not the only thing.

Until Ogg achieves significant success in most of the other areas discussed above, talking about quality is interesting, but a very small part of the much bigger picture that's yet to be painted.