Google promises publishers an alternative to AMP
Google's AMP project is not uncontroversial. Users often love it because it makes mobile sites load almost instantly. Publishers often hate it because they feel like they are giving Google too much control in return for better placement on its search pages. Now Google proposes to bring some of the lessons it learned from AMP to the web as a whole. Ideally, this means that users will profit from Google's efforts and see faster non-AMP sites across the web (and not just in their search engines).
Publishers, however, will once again have to adopt a whole new set of standards for their sites, but with this, Google is also giving them a new path to be included in the increasingly important Top Stories carousel on its mobile search results pages.
"Based on what we learned from AMP, we now feel ready to take the next step and work to support more instant-loading content not based on AMP technology in areas of Google Search designed for this, like the Top Stories carousel," AMP tech lead Malte Ubl writes today. "This content will need to follow a set of future web standards and meet a set of objective performance and user experience criteria to be eligible."
Also at Search Engine Land and The Verge.
Related: Kill Google AMP Before It Kills the Web
Google Acquires Relay Media to Convert Ordinary Web Pages to AMP Pages
Google Bringing Accelerated Mobile Pages to Email
(Score: 3, Informative) by Pino P on Tuesday March 13 2018, @04:15PM
True, Encrypted Media Extensions for digital restrictions management of web video require JavaScript. But that's not the only problem. The other problems are seeking, variable throughput of the Internet connection, and live video.
The naive method of seeking in a recorded video, relying on HTTP range requests, runs into two problems. The first is variable bit rate. Dropping the needle one-third of the way through a file won't get you one-third of the way into the runtime if the first third is encoded with a greater or lesser bitrate than the remainder. So a player has to use bisection search to figure out at which byte offset to start retrieving the video data, and this sort of back and forth can take a while over a high-latency satellite or cellular connection.
The other problem with HTTP range requests is web servers that fail to support range requests. Two decades ago, download managers used range requests to attempt to retrieve several pieces of a file at once, exploiting throttles that operated per connection. But because these used up more connection resources on download servers, several file download services deliberately disabled range requests except for those servers reserved for paying subscribers. A browser might discover that a range request has failed, and the server has fallen back to resending the entire video from the start.
Another cause of seeking is varying throughput of an Internet connection over short periods. To ensure a seamless experience for viewers, a service might want the player to choose among encodes at different bitrates. But when it switches bitrates, it has to seek to the corresponding position in the lower or higher bitrate encode.
Viewers expect to drop the needle in a live video at what's happening right now, or perhaps a minute ago ago after the video has passed through the buffers of state- or advertiser-required censorship and codecs that use B-frames. The naive solution of encoding the video separately for each viewer doesn't scale. An improvement is to encode once and start each stream at the next keyframe. But the architecture of widely used repeater services (also called content delivery networks) works on a whole file, not a stream that each server would need to demux and remux.
The common solution to these is breaking the videos into segments 3 to 10 seconds long, storing each at a separate URL, and linking them from a timed playlist. Then the CDNs can cache each segment, and the player can choose which to request based on the current playback position and recent throughput. But if the browser doesn't support common playlist formats, such as Apple HTTP Live Streaming or MPEG-DASH, a polyfill served by the website needs to handle this using the Media Source Extensions API, which requires JavaScript.
Not being a fan of porn, I don't know how long a typical porn video is. But if it's under a minute, the compressed video is likely to fit entirely within RAM, allowing the browser to use trivial seeking methods that rely on the whole muxed stream being available.