… the need to buffer the video before it starts playing will change the experience. Hence the experiment, rather than just a rapid rollout of this technology. On stage, he said the current resolution of YouTube videos has been “good enough” for the site until now.
I can only assume — or rather, hope — Chen is talking about genuinely high-quality video. Like ‘high def’ high quality. Because the issue with YouTube quality (apart from often shitty source material, which isn’t their problem) is that they’re using the video codec in Flash 5. This makes sense, because it means they can run ffmpeg on their server farm to transcode uploads, but there is a cost: Flash 5 video sucks.
The codec is Sorenson Spark, which is a close cousin of H.263 and H.264 but is visibly lacking compared to more modern implementations of the latter. Lacking, that is, at the same bitrates. There’s no shame in this, since Spark was designed years ago, for systems with far less compute power than can now be thrown at the challenge. But current H.264 implementations will deliver better quality and/or a larger frame and/or higher frame rate for the same compressed file size. And you tend to maintain audio sync, which is a particular weakness of Spark in Flash.
Higher-quality video needn’t mean lengthier buffer times. So… what was Chen talking about, again?
[Update: slightly more plausible report in the NewTeeVee Live official transcript. Still doesn’t make complete sense, though — Spark is old and shitty, they’re already doing better H.264 for iPhone and Apple TV… so… huh?]