There May Finally Be Better ATI Linux Video Playback

[bridgman]

See, when you say something like that it sounds like you're agreeing with Miles.

I don't understand. I am saying we need to make sure the DRM *doesn't* get hacked, and Miles is saying that we *want* to get hacked. They're about as opposite as you can get. Trust me when I say that our first response to finding that our DRM had been hacked would not be "ooh, let's open up even *more* information and see what *else* can go wrong"

I really hope you don't get hacked, because I'm guessing it would be bad for AMD and I don't want the limited OSS support we have now to dry up.

Yes, it would be bad for AMD and you can be pretty sure that open source support would dry up overnight. It might scare other companies enough to make them cut off open source support as well, but that would depend on the specifics.

Is anyone in the company nervous about this?

Yes, the risk of breaking DRM is a big concern, for us and for every HW vendor. This is why we stopped supporting open source development back in 2002, and why it was a big challenge to restart in 2007 and continues to be an expensive effort to maintain. The easy answer is not to open up UVD at all, but we are trying to do better than that. It's just a slow and expensive process.

Hackers can be a funny group, sometimes giving access even in a limited way can be enough to stave them off while an outright ban can draw attention. (See, PS3) NVidia allows people to use their hardware if they use the blob, AMD doesn't. Maybe that's enough of an argument to convince some skeptical suits?

Not sure about your last point. It should be pretty obvious by now that we *are* working on providing video accel support via the Catalyst driver... it just got leaked/discovered early so now you all get to watch how long development really takes instead of just being pleasantly surprised by a press release one day.

Providing open source support for UVD in order to stop hackers from reverse-engineering it would be like a bank giving away all its money in order to make sure it didn't get robbed.

[smitty3268]

I don't understand. I am saying we need to make sure the DRM *doesn't* get hacked, and Miles is saying that we *want* to get hacked. They're about as opposite as you can get. Trust me when I say that our first response to finding that our DRM had been hacked would not be "ooh, let's open up even *more* information and see what *else* can go wrong"

Yes, I know. But what you said was this:

to avoid releasing anything which might compromise the DRM on other OSes.

Meaning the problem is compromising the DRM on Windows. Meaning that if it was already compromised, that problem would no longer exist.

Not sure about your last point. It should be pretty obvious by now that we *are* working on providing video accel support via the Catalyst driver... it just got leaked/discovered early so now you all get to watch how long development really takes instead of just being pleasantly surprised by a press release one day.

I'm gonna be honest. It's NOT obvious. Maybe lots of stuff is going on behind the scenes, but from an end-user viewpoint it really looks like nothing is happening. When gbeauche discovers silly little bugs that would be a 1-liner to fix in the driver and nothing happens for 6 months leaving an entire generation of cards useless, I think it's fair to say that not much is happening. Or else the bureaucracy involved makes the federal government look like a well-oiled machine. I guess i had pretty much just assumed that AMD was happy with UVD support going only to embedded systems, or other paying customers, and that desktop linux users were going to be left in the cold. Are you officially saying that's not the case?

Providing open source support for UVD in order to stop hackers from reverse-engineering it would be like a bank giving away all its money in order to make sure it didn't get robbed.

I don't know, it depends on the design of the hardware i guess. It worked for Sony, though, because the PS3 only got hacked after trying to close it down.

[bridgman]

Yes, I know. But what you said was this: Meaning the problem is compromising the DRM on Windows. Meaning that if it was already compromised, that problem would no longer exist.

Um... yeah, like the problem of being depressed would no longer exist if you were dead. The immediate problem goes away but you don't necessarily have an improvement in the grand scheme of things.

I'm gonna be honest. It's NOT obvious. Maybe lots of stuff is going on behind the scenes, but from an end-user viewpoint it really looks like nothing is happening. When gbeauche discovers silly little bugs that would be a 1-liner to fix in the driver and nothing happens for 6 months leaving an entire generation of cards useless, I think it's fair to say that not much is happening. Or else the bureaucracy involved makes the federal government look like a well-oiled machine. I guess i had pretty much just assumed that AMD was happy with UVD support going only to embedded systems, or other paying customers, and that desktop linux users were going to be left in the cold. Are you officially saying that's not the case?

I can't agree with that at all. "Leaving an entire generation of cards useless" when those cards had never been supported in the first place seems like an odd complaint. I understand that users discovered the work-in-progress code, started using it while it was still in development, and felt that we were somehow obligated to make sure that unreleased code kept working for them and raise the priority to complete it on *their* schedule, but I don't think that is a fair expectation.

I'm not saying you don't have the right to complain about gaps in one vendor's feature set relative to others, but you have to admit that making assumptions about staffing level and priority then concluding that the teams are inefficient or incompetent based solely on those assumptions is a bit of a stretch.

I don't know, it depends on the design of the hardware i guess. It worked for Sony, though, because the PS3 only got hacked after trying to close it down.

AFAIK the PS3 had been reverse engineered for a long time, and it was only the new lock that got hacked after trying to close it down.

[smitty3268]

I can't agree with that at all. "Leaving an entire generation of cards useless" when those cards had never been supported in the first place seems like an odd complaint.

Unsupported == Useless, at least in my book. I'm not sure why anyone would think differently. As far as UVD, i mean. Obviously if users weren't interested in UVD then the other functionality was present.

but you have to admit that making assumptions about staffing level and priority then concluding that the teams are inefficient or incompetent based solely on those assumptions is a bit of a stretch.

I never assumed they were incompetent. I assumed that the suits in charge had made a business decision that UVD support would only be provided to customers who paid for it, and then only for the specific cards they paid for the support. I don't think this is a stretch at all - I still don't think i've ever seen a single quote from AMD saying otherwise.

AFAIK the PS3 had been reverse engineered for a long time, and it was only the new lock that got hacked after trying to close it down.

No - there was some USB flaw a while ago that got hacked, but that was fixed by later firmware updates and what you could do with it was of limited scope. The current situation is completely different - they've basically cracked every layer of security Sony put into the PS3, down to getting the official private keys they use to sign applications. They've already created their own custom firmware to replace the official Sony version, now.

[bridgman]

Unsupported == Useless, at least in my book. I'm not sure why anyone would think differently. As far as UVD, i mean. Obviously if users weren't interested in UVD then the other functionality was present.

What I'm trying to say is that if you are using unreleased code and it stops working you don't have the same moral right to "justified outrage" as you do once development is finished and the feature is officially supported.

I never assumed they were incompetent.

Sorry, perhaps I misinterpreted your words :

When gbeauche discovers silly little bugs that would be a 1-liner to fix in the driver and nothing happens for 6 months leaving an entire generation of cards useless, I think it's fair to say that not much is happening. Or else the bureaucracy involved makes the federal government look like a well-oiled machine.

You seemed to be assuming that the developers priorities were the same as yours (making unreleased code work for unsupported boards, because you had already started using it) and concluding that nothing was happening or that the work was being badly managed. Sorry if I got that wrong.

I assumed that the suits in charge had made a business decision that UVD support would only be provided to customers who paid for it, and then only for the specific cards they paid for the support. I don't think this is a stretch at all - I still don't think i've ever seen a single quote from AMD saying otherwise.

But... but... but... we don't pre-announce other new features, why would you expect video decoding to be any different ?

No - there was some USB flaw a while ago that got hacked, but that was fixed by later firmware updates and what you could do with it was of limited scope. The current situation is completely different - they've basically cracked every layer of security Sony put into the PS3, down to getting the official private keys they use to sign applications. They've already created their own custom firmware to replace the official Sony version, now.

Ahh, OK. I had a different understanding of the history there. Thanks.

[smitty3268]

What I'm trying to say is that if you are using unreleased code and it stops working you don't have the same moral right to "justified outrage" as you do once development is finished and the feature is officially supported.

Fair enough.

You seemed to be assuming that the developers priorities were the same as yours (making unreleased code work for unsupported boards, because you had already started using it) and concluding that nothing was happening or that the work was being badly managed. Sorry if I got that wrong.

Yeah, what i really meant was that the developers at AMD are too good to simply be unable to fix this problem. Which means that the reason it's not getting done is coming from above somewhere. The beancounters, or the lawyers, or some kind of hellish bureaucracy. My assumption was the beancounters, since you had stuff working in the drivers and were only allowing certain people to actually use it. I also assumed that the UVD work was probably pretty specialized, and wouldn't have people taken off of it to work on some unrelated issue like a 3d GL bug - maybe that's wrong.

But... but... but... we don't pre-announce other new features, why would you expect video decoding to be any different ?

Because this is pretty much a core feature of the graphics card and one that all your competition has but you don't. I don't expect any dates, but a "yes we're working on this" would be good. I haven't seen that. If you guys weren't supporting OpenGL 4 yet, or DX11, I'd also be wondering why you hadn't announced that you were working on it yet, and if you planned to ignore it entirely. Like it or not, if every competitor has an important feature except for you and you refuse to talk about it, people are going to wonder what's up.

[RealNC]

No one's interested in cracking AMD's DRM because it doesn't matter. It really isn't good for anything else than playing protected media. It does not prevent the playback of unprotected content (like pirated videos). So it protects nothing: If you're using protected media, then you paid for it. If you're using unprotected, pirated media, then you didn't pay for it, but can still watch it. That's the very definition of "useless."

In other words, it's totally useless. Making a fuss about it makes them look silly, IMO (being careful not to release something that can lead to cracking something that doesn't need cracking in the first place because it's a useless piece of crap to begin with.)

[not.sure]

The insanity of DRM is hardly their (AMD's) fault. They have to do what they have to do.

[bridgman]

Yeah, what i really meant was that the developers at AMD are too good to simply be unable to fix this problem. Which means that the reason it's not getting done is coming from above somewhere. The beancounters, or the lawyers, or some kind of hellish bureaucracy. My assumption was the beancounters, since you had stuff working in the drivers and were only allowing certain people to actually use it. I also assumed that the UVD work was probably pretty specialized, and wouldn't have people taken off of it to work on some unrelated issue like a 3d GL bug - maybe that's wrong.

Yep, UVD work is pretty specialized, but that only means that the UVD team is dealing with 10-15 different projects instead of hundreds. The main reason developers aren't working on project A (the one you want), or aren't working as fast as you would like, is that they are also (or instead) working on projects B, C, D, E and F, all of which are exquisitely important for some of our customers. Nobody sits idle, and nobody is hamstrung by bureaucracy, but they may be working on things that you don't care about but other customers do (and, predictably, those customers are wondering why the heck we work on Linux).

Because this is pretty much a core feature of the graphics card and one that all your competition has but you don't. I don't expect any dates, but a "yes we're working on this" would be good. I haven't seen that. If you guys weren't supporting OpenGL 4 yet, or DX11, I'd also be wondering why you hadn't announced that you were working on it yet, and if you planned to ignore it entirely. Like it or not, if every competitor has an important feature except for you and you refuse to talk about it, people are going to wonder what's up.

Yep, that's fair, but everyone in this business has learned that if you announce your roadmaps you make more people unhappy when plans change than you made happy by announcing the roadmap in the first place... so nobody pre-announces and that's just the way it is. When we had OpenCL and other vendors didn't, nobody pre-announced, same for EyeFinity etc...

[Qaridarium]

Providing open source support for UVD in order to stop hackers from reverse-engineering it would be like a bank giving away all its money in order to make sure it didn't get robbed.

you are maybe wrong here

http://en.wikipedia.org/wiki/W%C3%B6...rgl_Experiment