3 ms·
The idea of a 'write-only layer' in the browser scenarios described is ridiculous. It simply doesn't exist. In the days of DVDs, GPU acceleration exposed the c
by kevingadd 13y ago
The idea of a 'write-only layer' in the browser scenarios described is ridiculous. It simply doesn't exist.
In the days of DVDs, GPU acceleration exposed the concept of 'overlays' that you could dump YUV or RGB video content into, and those overlays would be presented directly to the display - the RAMDAC overlaid them onto your framebuffer when producing VGA/DVI/etc signals. This, at least, made it possible to make such a mechanism 'write only', if you sufficiently secured the output (via negotiation, or something) to ensure it wasn't being fed to a capture device - you only had to deal with the analog hole.
But once you're trying to composite protected content into a webpage, you've already lost the game. The content can be clipped, scaled, rotated, sheared, faded, masked, filtered, and so on - all without writing any JS. All those transformations would somehow have to be baked directly into this 'protected rendering' pipeline and implemented cross-platform in such a way that none of that data was accessible at any point in its myriad trips between the CPU and GPU. You'd have to access-control all the cache textures on the GPU, all the buffers on the CPU, all the intermediate information. You'd also have to access-control the resulting framebuffer for the browser tab, and any other intermediate framebuffers, along with the actual output from the GPU that's sent to your display - and most displays these days are connected digitally, so now you're dependent on your end-users having HDCP compliant displays (which many do not).
It's a complete joke. EME cannot work. Period. Modern GPU/CPU configurations are basically general-purpose compute hardware with mutually accessible memory and the ability to do computations and read/write operations from inside rendering routines (i.e. pixel shaders).
- acdha 13y ago> The idea of a 'write-only layer' in the browser scenarios described is ridiculous. It simply doesn't exist. I'm sure that will be quite a surprise to Microsoft as they shipped it years ago. You couldn't push every possible operation out to the GPU but there's no technical reason why the subset of operations which a browser would use to playback a video couldn't be supported, similar to the way a modern desktop can be GPU accelerated but e.g. screen captures don't include windows where restricted content is playing. The problem is not technical – that ship sailed years ago – but business: the content providers and consumers have accepted the current status quo with plugins rendering the subset of paid content which isn't playing in a native app. The businesses don't need perfection – they're already making truckloads of cash. As I stated above, we really need to figure out how to disrupt that if you want to see any change. Pretending that technical perfection is necessary or that the public isn't willing to tolerate a billion dollar a year status quo is just delaying that discussion.
- shmerl 13y ago> the content providers and consumers have accepted the current status quo with plugins rendering the subset of paid content which isn't playing in a native app. This completely misses the point. Many users didn't accept any of that and simply pirate all that content DRM free. If publishers care about increasing the number of paying customers, they can stop using DRM, and then some part of former pirates will start paying them. They are usually greedy for more money, so "already making trackloads of cash" doesn't sound like an argument.