Vimeo tech chief takes on 'terrifying' online video challenges (Q&A)

Think online video is old hat? Think again.

It was eight years ago that Google bought YouTube for $1.65 billion, so you'd be forgiven for considering that streaming video has become a routine part of the technology world. As chief technology officer of Vimeo, though, Andrew Pile faces an unending series of challenges making sure his site works today and tomorrow.

"It has been terrifying at times," Pile says of trying to keep up with online video technology.

He's got to wrestle with the changing way data is disseminated over the Internet and the tension between mobile apps and the Web. He's had to step into the middle of surprisingly emotional technology debates over things like video compression formats, copy protection and Adobe's Flash Player browser plugin.

It may not be your job, but you might well care about how well Pile does his, because you're likely someone who uses the site. Vimeo had about 55 million visitors in September in the US alone, according to traffic analysis site ComScore, and the company revealed in July that people upload 500 hours of video each hour. Technology problems can mean that viral video your friend shared on Facebook doesn't work, that short film stutters or looks pixelated, or that movie trailer you're watching stops playing halfway through.

Pile talked with CNET staff writer Stephen Shankland about the company's technology challenges and how he's addressing them with changes like a new mobile Web app. The following is an edited transcript.

Q: With the HTML5 standard, video now can be built into the Web, and Vimeo was an early fan compared to Flash. What's your split now between HTML and Flash for videos delivered?
Pile: The lion's share is HTML5. It's got pretty decent penetration. We're still delivering Flash to Facebook, because their feeds don't allow for framing of HTML content.

So the number of Web browsers that can't handle HTML video is pretty low?
Pile: We found support to be pretty good. For about a year now, we've had nice stability.

For a while, the Web was the future for programmers trying to reach customers, but in recent years, there's been a huge pull toward mobile apps. How many people use the Web version of Vimeo and how many use the mobile app?
Pile: It's a large app, but it doesn't compare to the amount of traffic we get on the mobile Web. Anyone who is linked to from a feed [like the constant stream of updates people see with Facebook and Twitter] experiences a huge amount of mobile Web traffic.

Mobile is now a big percentage of our onsite traffic. It's now over half of our traffic in terms of unique users.

What's the long-term trend for mobile apps vs Web access to Vimeo?
Pile: If you're a serious Vimeo user, you want to browser more easily, search, and delve more deeply, then the app experience makes more sense. [For just following a shared link] I'm not sure people want an app. They want the convenience of not having to leave Twitter or Facebook to do that.

We're working on a new version of the mobile website. We call it drive-by consumption. Today, people see a video in a feed, click the link, go to Vimeo to view it, then click out. We'd like to get that more compelling so people watch other pieces of content or share it.

There are challenges we are taking on. On Android [Google's mobile operating system] it can be different, but on iOS [Apple's mobile OS] you can't be logged into Twitter and Vimeo at once. A huge percentage of our traffic is sandboxed inside other people's apps like Twitter.

When will that new mobile website arrive?
Pile: Before the end of the year

One tough part of the transition from Flash to the Web has been with digital rights management (DRM). A lot of people, including some writing the HTML standard, don't like copy protection, which can be pretty user-hostile. But Hollywood likes it, and there's an increasing amount of premium video going online. HBO and CBS are the latest to embrace it. What's your take on Encrypted Media Extensions for HTML video?
Pile: We don't use it yet. We have our Vimeo on Demand service, where people can put up their own work for sale, but people producing the types of content we're doing, DRM has not been a huge issue. They're more interested in getting audience than protecting content per se.

But as we get to scale, people are going to start caring more and more. I think it's good we'll be able to use HTML5 to deliver this in a secure manner. I prefer that to being locked in a proprietary solution that limits playback in certain places. That said, it is somewhat contrary to the open Web. If I were at a different organization, I could be more high-and-mighty about it, but for us it is probably a good move.

So does Vimeo have ambitions to let people watch the standard sort of premium video -- the latest movie from Sony Pictures?
Pile: I think so. As DRM becomes more [of a] requirement for our Vimeo on Demand product, we're going to have to be moving this way. We've been holding out because it's a lot of work to implement.

Mobile apps are a big deal, but there are two major operating systems, Google's Android and Apple's iOS, and then there's also Microsoft's Windows Phone and more. Does the Web help you reach other platforms without having to rewrite the same apps over and over?
Pile: If we don't have an app on these platforms, there is that safety net. If I'm using a Firefox OS phone, I can still see Vimeo content and links work. For us the primary use case is fulfilled. The secondary use case is upload, and that's not always there.

In mobile, there are a couple kingdoms. But if you want to say you work on smart TVs you have have something like six apps. Set-top boxes have become powerful, they have great user interfaces, they're updated, and they have programming interfaces to do cool things on systems like Roku, Apple TV, Amazon Fire and Microsoft Xbox. Some winners are starting to emerge, but it's not at all set in stone. You need be everywhere for it to be valuable.

People are starting to realize companies aren't going to spend tens of thousands of dollars writing apps for a platform that has no user base and might be obsolete in a year. So people are starting to adopt more of an HTML approach.

The H.264 video codec is dominant right now, but HEVC/H.265 is on the way as a better way to compress video. Have you tried it yet?
Pile: Yes, we experiment with all that stuff and try to understand it and be prepared for what's coming. There's a huge gap between what's possible and what we do. We're pretty bullish on H.265, but it doesn't have the amount of penetration we need for it to be worthwhile yet. For the 2K and 4K [high-resolution video] range we'll be doing H.265. That's the realm where more efficiency [higher video resolution for a given amount of network capacity] is key.

Have you thought about serving H.265 so you could save bandwidth on low-resolution videos, too?
Pile: For us, the cost centers are not just delivery, but also storage. We might get more efficiency in streaming, but we have to store that file too.

How about real-time transcoding, where you store a master version and create other formats only when you start streaming?
Pile: For [low-resolution] SD and mobile video, we're looking at that pretty aggressively. We'll start at the low end and work our way up.

How much storage do you have?
Pile: It's many petabytes. We store source files and transcoded files in different places and different ways.

So supporting Google's VP8, VP9, or VP10 format would mean a lot more storage? Lots of mobile chips now support VP8, and VP9 is built into Chrome and Firefox now.
Pile: The question is how you get better client support without drastically increasing your storage costs. There are things you can do with transcoding or repackaging in real time. That's stuff we're experimenting with.

You like the H.264 compression format. How did you feel when Google announced in 2011 that it was going to drop H.264 support from Chrome -- a decision it later abandoned, of course?
Pile: That was definitely concerning. In the worst-case scenario, we could fall back to Flash. We have done that for various releases in the past where browsers sometimes do wonky things.

What's your opinion on Google's push to promote its royalty-free VP video technology? There's quite a religious war pitting it against H.264.
Pile: They're great codecs. I think it's awesome they want to release a new version every year. Google is a powerhouse when it comes to that. But it's also telling that only [Google-owned] YouTube has adopted it. For us, storage cost is an issue, but for them bandwidth is an issue. It's critical for us, but it's at a different scale. With H.264, the video file is going to work on your phone, your TV, your computer. Unfortunately, the VP online of codecs is so far away from that universe that we just haven't gone there.

But the purchase of [VP developer] On2 Technologies by Google really did force the H.264 [patent issue]. I wither in any sort of patent talk. It's still terrifying. It did say a lot for how much Google cared about video being an intrinsic part of browsers. It did help with unknowns of patents, and the licensing around H.264 was terrifying.

MPEG LA, which licenses the H.264 and HEVC/H.265 patents, said H.265 will be royalty-free for video streaming, including for paid video services like Vimeo on Demand or Netflix.
Pile: I doubt that would have happened without Google.

Another idea from Google, the SPDY network technology, has caught on more widely in browsers. But it requires end-to-end encryption from the server all the way to the browsers. That makes life harder for the content delivery networks [CDNs] you and others use to distribute content all over the world. What do you think of SPDY and end-to-end encryption?
Pile: That's an extremely pointed question. The entire world is going to HTTPS [encrypted Web connections]. Our site uses HTTPS if you're logged in, but it's [unencrypted] HTTP if you're not. We're changing that as fast as we can. Everything needs to be secure.

When all these big CDNs were architected, HTTPS security was a separate product. The way it's going is all one product. Getting vendors to align with that has been more difficult than we'd like. Some see the green lock in the browser address bar as vow of secrecy -- everything from the original server to browser is secure. But some people just see it as a hurdle to be overcome.

How does gigabit broadband affect your business?
Pile: We're more than happy to deliver the highest quality we can push down anyone's connection. If the bandwidth is there, we would love to use it. [However, there are limits to] the insight we have into consumers. It's really hard to diagnose if there's a problem, especially if we're working with a [content distribution] vendor in another country.

Can you use adaptive streaming, so the resolution of the video is adjusted on the fly?
Pile: That's something we're heavily working on, but it's a ways off. YouTube's implementation is really spectacular. But they have the browser and they write their own video compression standard. They're able to do a tight loop so they can do this awesome implementation. It's a requirement for them operating at the scale they're operating at. We want to get to that place, too.

Has video on the Web been frustrating or delightful? The technology is important, but it changes so fast.
Pile: It goes in fits and starts. Back in 2005, 2006, 2007, when Flash video started supporting H.264, it allowed us to do HD video better. Then there were been periods with VP8 vs. H.264 and Google saying they were going to pull support for it -- that was frightening. A lot of the time we're in the passenger seat when it comes to these macro movements of video online. Now we've gotten big enough it's more of a pleasure than it has been. Online video is such an intrinsic part of people's experience.