The Linux Kernel Begins Preparing For AV1 Decode Support

[chithanh]

Sisvel's conduct suggests that their intent is to harm the adoption of AV1; not to get licensing money out of it, and not to get specific infringers to cease.

Sisvel may never bring a claim in court.

Even before that, if they ever try to extract royalty payments out of any Chinese company, they will have to reveal the specific relevant patents. China ensured this after Microsoft made Android vendors pay for their patents on Linux. (What happened next is that independent review of the patent claims could start, few months later Samsung stopped royalty payments, Microsoft sued but ultimately was left with only getting Office apps preinstalled on Samsung Android devices, others followed suit).

Yeah, might not be worth fighting for years. You Do know, that Google claimed VP8 is not infringing (not sure if thats a word) on Mpeg LA patents, then got the license they dont infringe on years later (likely paid for it)?

There was only a token one-time payment, it suggests that MPEG LA had a very weak case, and it was more a payment to make the inconvienience of prolonged lawsuits go away, and MPEG LA could save some face.

[microcode]

IF we go conspiracy mode, this is more like swaying the public opinion for their own format, instead of being upfront about patents.

Thanks for sharing the list, I guess this was published finally this year. Has anyone done a review of these to see which ones are the most credible, and evaluate which tools they affect?

Seems like the best response, if practical, is to simply avoid the credible claims (if this is practical).

[OneTimeShot]

Google paid a bit of money for VP8 a long time ago (10 years) when there were still some valid video patents around. Patents only last for 20 years, so even the original h264 patents will expire in the next couple of years.

The days of claiming patents on video technology is basically over. In any case, the primary inventors these days are YouTube and Netflix - so the shoe is very much on the other foot as to who is infringing who's IP.

[mdedetrich]

Nope, it's duplication, like v4l2.

Partly because reinventing the wheel is what ALWAYS happens with anything to do with media (insert links to any of a dozen xkcd strips here :P), but also partly because VAAPI has some pettiness in it that guarantees you can't actually have a conformant implementation unless the hardware is built in a particular way / has certain capabilities that have nothing at all to do with video.

Uh what, how does this make sense?

[arQon]

Uh what, how does this make sense?

Because VAAPI was written to match the capabilities of Intel's IGP of the time. Since Broadcom half-assed the SoC the way they did, the video block can't do ANYTHING except decode. (And it's also stateless-only).
I can't remember now if the issue is that VAAPI requires that the same device is also capable of 2D ops, or if it's the stateless part that breaks it (there was a lot to dig up and make sense of, with none of it being particularly well documented), but the end result is the same: no VAAPI, and no possibility of it, ever.