Not true at all, there is nothing in the spec that mandates the use of any particular OS.

But there's nothing compelling plugin developers to support all OS/browser combinations, so no, in practice it likely won't be OS-agnostic.

This scheme basically just creates a special class of plugins; these plugins clearly won't be OS-agnostic, because they can't -- that's the whole point of the exercise: to restrict playback to devices that are fully authorised/controlled from top to bottom, with the browser piping streams from the web to trusted plugins running on trusted OSes using trusted hardware (TPM etc).

Isn't this the same with all software? There is never anything mandating that anybody supports your OS or browser of choice.

If I want to watch netflix or play GTA5 on my haiku box I'm SOL as it is unless there is a business case to be made for doing the port.

This actually makes things easier. For example netflix currently uses silverlight for their streaming, this means that in order to watch netflix you need something that supports the entire silverlight stack.

With this proposal all you need is modern browser and a compatible CDM which is a much smaller chunk of code.

Much smaller, I don't think. It will still have to grab the screen and render directly onto it. It's basically the same thing a regular plugin would do except a little bit of identity validation, from what I understand.

Technically speaking it's almost the same as the status quo, but a little bit of DRM principles have now been enshrined in the foundations of our web. Depending on what your view of DRM is, of course, this is good news or bad news; but from a technical point of view, it changes very, very little.

Indeed, this changes nothing technically. But what it does is harm the idea that standards should promote the the open web. You may or may not agree to this idea of course.

> Isn't this the same with all software?

That's exactly the point, the case is NOT so with HTML and JavaScript. A webapp I write in HTML and JavaScript is OS agnostic!

It's not the same because HTML5 EME is designed to ensure anyone who manages to get media working on unauthorized platforms can be arrested under the anti-circumvention provisions of the DMCA and other similar laws. As in, this was actually one of the stated requirements.

This is simply incorrect, can you show a w3 document that supports this in any way?

Is there anything in the spec that ensures compatibility with more than one OS?

Is there anything in the TCP/IP spec that ensures compatibility with more than one OS?

TCP/IP is an open standard, DRM are not.

Your statement is as wrong technically as stating that encryption cannot be open source. What matters are the keys, not the code.

The purpose of encryption is to secure communications between two trusted endpoints against untrusted intermediaries. Since all parties to the exchange are trusted and the intent is for both parties to have access to the data, an open source encryption system works.

DRM's purpose is similar, but one of the endpoints is in the physical control of an untrusted entity: you. Since the point of DRM is to prevent the user at one of the endpoints from accessing the data, if you have the source and keys to the destination endpoint (e.g. TPM, HDCP-enabled GPU), the endpoint can't be trusted, you can get the data, and the DRM fails at its purpose.