While the author is correct, the tone feels so wrong. It's terrible that laws exist that make knowing certain information a liability to embark on creative endeavors. Just a very real view on why IP laws are harmful (despite potential upsides).
Also, what are the kind of super secrets in drivers? If the secrets are so valuable, certainly competitors would have no trouble simply getting a spy to work as an employee. Or just RE from the binaries.
It reminds me of the idiocy of Canon and Nikon in refusing to document their raw formats.
The secrets contained in drivers and firmware aren't valuable, merely dangerous. Having the source to a graphics driver doesn't make it much easier to design graphics hardware, but it does make it a lot easier to identify and prove patent infringement, and it makes it easier to circumvent arbitrary restrictions that exist for purposes like product segmentation or DRM.
> It's terrible that laws exist that make knowing certain information a liability to embark on creative endeavors.
Why?
We don't find anything objectionable about the law extending property rights to, say, land, which we had no hand in creating and will continue to exist long after we're all dead, so why find an objection to Imagination having property rights over something that is wholly their creation?
I would argue that to the extent you can morally justify any sort of property rights, copyright and trade secret more more justifiable than nearly any other kind.
Furthermore, I don't see what creative endeavor is being harmed here. It reduces the meaning of creative to nothing if you apply it to reverse engineering efforts, which are the opposite of creative--they are derivative.
Private property is a system that's useful to reduce conflicts that arise from two properties of land: rivalry and scarcity. Ideas and other creations need to such system, since they can be shared infinitely and concurrently.
Also, the "wholly their creation" is often untrue, especially when it comes to patents. It's often just a result of having the money and/or knowledge to submit it. Multiple people can and have came up with the same patentable concept independently.
We consider it acceptable for someone to own a thousand acres of ranch land that they only visit in the summer. Clearly our understanding is not based purely on the rival nature of land.
Implementing an API can still be creative. I wasn't allowed to work on Mono, as I had MS source code access. Just on virtue of having perused MS's source, I couldn't participate. I understand the reasoning, but the end result feels wrong.
Land ownership and IP are completely different, for example because nobody ever, and fortunately it'll forever impossible, paied property taxes on ideas/discoveries/thoughts.
Actually I like that idea. If patent holders were required to state an expected value for their claims and had to pay property taxes on that expected value every year, we'd see far fewer stupid patents held by unproductive trolls. (And fewer outlandish damage awards since the expected value could be cited in court.)
I doubt that IMG will now try to bullshit us with the inane patent excuse.
It's not inane. Their fear is a natural consequence of mixing ludicrous patent systems with cutthroat industry politics.
Most GPU vendors don't have massive patent war chests of their own. They have a lot to lose by exposing their implementations, because of course they all infringe on various patents held by other vendors. Any program longer than hello.c is likely to infringe on various patents. When there's money at stake, the incentives all point towards keeping the driver sources private.
It is not other vendors that you need to be worried about. For example, when joining Khronos (OpenGL, GLES and others) you get royalty free permission to use any technology in Khronos' specifications and you, as a vendor, have to offer your tech to other Khronos members for free. I wasn't involved in the DirectX side of things but I imagine Microsoft operate a similar IP framework. There is no practical way to be a GPU vendor today and hoard your IP.
Patent trolls though. A different story. In 5 years we were sued twice (that I know of) by trolls. On both occasions the trolls had some trivial enhancement to a well known technique. I imagine they are operating a mud slinging operation where eventually something sticks. Once it does then either (1) the cost of licensing this 'new' tech is priced less than legal costs or (2) the troll is looking for testimony from some respected entity that the tech is innovative which gives them validation to go after the next biggest vendor up the chain.
> Most GPU vendors don't have massive patent war chests of their own.
The largest by marketshare, nVidia and ATI (by belonging to AMD) certainly have a sh.tload of patents. Do correct me, but aren't IMG/PowerVR and nVidia holding the majority of the mobile GPU market share anyways?
NVidia is a minor player in the non-laptop mobile space. PowerVR and Qualcomm have the most popular GPUs, followed by Vivante and ARM's Mali as the second tier. NVidia's mobile GPUs are only used in their own SoCs, and the Tegra has never been that successful.
Sorry, but it really just does not work that way in the patent world (copyright is a different story, but they have never made the claim they are worried about copyright infringement).
I can't even remember a case that started by someone looking at source code to figure out if their IP was being used.
I take the exact opposite stance, since the practice of secretly studying "leaked" proprietary code is not new - one of the earliest examples I can think of is the Lions' book[1]. In fact, knowing how the proprietary version does it could be valuable in writing your own implementation, as then you can be certain you're not coincidentally doing the same thing and use an approach which could be better in some ways (more efficient, etc.)
Given one reason gpu manufacturers use for keeping drivers closed is to avoid patent cases/trolls, can prosecution lawyers use leaked IP as a basis for a patent case?
Would that then lead to a counter-suit for illegal ip access?
"Given one reason gpu manufacturers use for keeping drivers closed is to avoid patent cases/trolls,"
No, this is what they state. IT is of course, not in any way accurate.
Having access to source code does nothing for patent trolls, or anyone else.
They don't go scouring source code and then suing people.
They go suing people and then scouring source code.
Since they'd get source code access in discovery anyway, it doesn't matter whether they publish or not.
It's really that they think they have magic secret sauce.
I don't get what they author want to tell me. This not great for powervr because now their chips will be cloned and competitors have access to trade secrets.
Neither of that will get worse by some non commercial developers looking at the code. Nor will their code become a copyright infringement unless they do copy and paste. And with a VCS (that everybody with a task that big needs anyway) it's not hard, if trivial, to show it's not. And ever better, they don't need to show it's not, somebody else needs to show it is.
Source to look at is a great thing for everybody who tries to understand powervr. Not less work, but more information.
"I am horrified about the lack of responsibility of a lot of people. These are not some cat pictures, or some nude celebrities. This is code that forbids people from writing graphics drivers."
I'm not sure if that line is mean to be sarcastic but if taking literally definitely undermines the credibility of the writer !
If we weren't able to narrow our focus, then we'd be all wound up about everything that's supposed to be an outrage and we'd never get anything done. How many outrages? How many people you got?
Also, what are the kind of super secrets in drivers? If the secrets are so valuable, certainly competitors would have no trouble simply getting a spy to work as an employee. Or just RE from the binaries.
It reminds me of the idiocy of Canon and Nikon in refusing to document their raw formats.