Announcement

Collapse
No announcement yet.

Nouveau: NVIDIA's New Hardware Is "VERY Open-Source Unfriendly"

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts
    chithanh
    Senior Member

  • chithanh
    replied
    No, I think the message is: The FSF will recommend your hardware as long as it can be properly supported by free distros.
    Whether you wrote the driver yourself, or someone else did this for you, is less important.

    Originally posted by bridgman View Post
    It's hard to see how that position does anything but discourage future investment from HW vendors, in fact it makes "doing nothing" seem like the winning strategy. Are you sure that's the message we all want to be sending ?
    I don't think any significant number of users would use Linux on Radeon graphics hardware today, if AMD had done nothing. Those who did would have been stuck with xf86-video-avivo and functionality comparable to the gma500 driver (or Catalyst and old kernels/X servers). ATI presence was not particularly strong on Linux, so the interest to reverse engineer their hardware/drivers would have been limited.

    And I believe (but don't have proof, so feel free to disagree) that your employer totally underestimates the importance of winning the minds of the Linux nerds, who are often relied on in hardware purchasing questions. Poor Linux drivers (and the Catalyst Linux driver is deficient in many aspects) would have hurt AMD's Windows sales too.

    Leave a comment:

  • bridgman
    AMD Linux

  • bridgman
    replied
    Stupid edit limit.

    Originally posted by chithanh View Post
    What was the motivation to write a free replacement for the NVidia firmware is inconsequential for inclusion in a free distro. Yes, and even RMS deplores that he has to recommend NVidia hardware, a company he calls hostile to free software.
    It's worse than that. FSF's position sends a strongly discouraging message to other vendors thinking about supporting free drivers for their hardware.

    My point about the motivation was that the arrival of RE'd microcode for NVidia HW ended up leading FSF to a position which seemed convenient at the time but which directly conflicts with almost everything they believe in and are trying to accomplish... and that it was only the combination of redistribution problems and "having nothing to lose" (ie no support from the HW vendor) that drove the community to invest in RE'ing.

    It's hard to see how that position does anything but discourage future investment from HW vendors, in fact it makes "doing nothing" seem like the winning strategy. Are you sure that's the message we all want to be sending ?
    bridgman
    AMD Linux
    Last edited by bridgman; 02 May 2015, 03:12 PM.

    Leave a comment:

  • chithanh
    Senior Member

  • chithanh
    replied
    They don't, but people RE'd it anyway. Of course this can get anybody who distributes the code into legal trouble, which was why e.g. the acx100 driver was denied kernel inclusion.

    Leave a comment:

  • bridgman
    AMD Linux

  • bridgman
    replied
    Originally posted by chithanh View Post
    What was the motivation to write a free replacement for the NVidia firmware is inconsequential for inclusion in a free distro. Yes, and even RMS deplores that he has to recommend NVidia hardware, a company he calls hostile to free software.

    Besides, the radeon microcode license forbids reverse engineering, so the only way someone can ethically use proprietary software according to RMS (namely, in order to write a free replacement for it) is legally blocked by AMD. Some people are working on reverse engineering Adreno microcode though, which seems to be quite similar to R600 microcode. So maybe in the future RMS can recommend older AMD hardware.
    I'm sorry, I missed the part where NVidia and Qualcomm licenses allowed reverse engineering. Can you point out the relevant text please ?

    Leave a comment:

  • chithanh
    Senior Member

  • chithanh
    replied
    Originally posted by bridgman View Post
    The NVidia microcode wasn't RE'ed because of an overwhelming desire for freedom, it was RE'ed because NVidia did not provide the microcode with a license allowing redistribution. On the other hand I spent time with distro & AMD legal folks to work up a license that *did* meet redistribution requirements, so NVidia HW gets recommended and ours does not.
    What was the motivation to write a free replacement for the NVidia firmware is inconsequential for inclusion in a free distro. Yes, and even RMS deplores that he has to recommend NVidia hardware, a company he calls hostile to free software.

    Besides, the radeon microcode license forbids reverse engineering, so the only way someone can ethically use proprietary software according to RMS (namely, in order to write a free replacement for it) is legally blocked by AMD. Some people are working on reverse engineering Adreno microcode though, which seems to be quite similar to R600 microcode. So maybe in the future RMS can recommend older AMD hardware.

    Originally posted by bridgman View Post
    Yes, but your explanation seemed to be a circular argument, basically "it's OK because the FSF-approved distros say so".
    No, I explained why firmware in flash is ok in the sense that you could then run free distros with it. Not what I personally consider ok for myself.

    Leave a comment:

  • bridgman
    AMD Linux

  • bridgman
    replied
    Originally posted by chithanh View Post
    Of course you can wait for the FSF to change their stance on proprietary software. But I wouldn't hold my breath. It appears that you need to accept the fact that their ethical views cannot be reconciled with yours.
    I accepted that a long time ago -- otherwise I would have had to walk away from the open source effort.

    These days it's just the sheer illogic of the situation that offends me

    Originally posted by chithanh View Post
    I think I explained this in the post you replied to.
    Yes, but your explanation seemed to be a circular argument, basically "it's OK because the FSF-approved distros say so".
    bridgman
    AMD Linux
    Last edited by bridgman; 02 May 2015, 02:47 PM.

    Leave a comment:

  • chithanh
    Senior Member

  • chithanh
    replied
    Originally posted by bridgman View Post
    ... and if you buy into the (IMO mistaken) belief that binary microcode distributed with the hardware is somehow not a problem, even though the same binary microcode distributed other ways (even with controls on updating) is somehow evil.
    The microcode is software, and it is objectively not free software. It can thus never be included in a distribution which is restricted to shipping only free software. Which is totally orthogonal to the question whether proprietary software is evil.

    Originally posted by bridgman View Post
    Another circular argument. You're assuming the rules used to create FSF-approved distros are "right" - I'm arguing that they are not only wrong but are sending a message to users that even the FSF is not happy with (although so far they have not yet been sufficiently unhappy to change it).
    I am just relaying what I think is the FSF's position. I wrote to you before that you are welcome to have a low opinion of that position.

    Originally posted by bridgman View Post
    The same would be true if the rules used to create FSF-approved distros were updated -- what I would call "fixed"
    Of course you can wait for the FSF to change their stance on proprietary software. But I wouldn't hold my breath. It appears that you need to accept the fact that their ethical views cannot be reconciled with yours.

    Originally posted by bridgman View Post
    And still we have the case where upgradeable microcode in flash is OK but the same microcode in a file is not.
    I think I explained this in the post you replied to.

    Leave a comment:

  • bridgman
    AMD Linux

  • bridgman
    replied
    Originally posted by SystemCrasher View Post
    I can see only one valid reason to dislike uCode on disk more than one in flash. If uCode requires special licensing to redistribute it, it poses trouble for, say, some new distro who can't easily obtain permission to redistribute foreign binary. This sucks and it could be one of valid FSF concerns. Because distro would be crippled in terms of hardware support compared to some more lucky entities who got such permission.

    But as long as vendor permits unlimited distribution of code, it is not anyhow worse than some pre-existing code in ROM/flash/etc. So I guess all this zealotry haves little to do with AMD cards (sure, opensource code could be better, but at least, as long as AMD permits distribution and use of their ucodes. they're not anyhow worse than if they were stored in flash).
    Yep. I keep coming back to the conclusion that there was a necessary clean-up because a lot of the microcode did not have licenses with appropriate redistribution rights, and that we just got caught up in the enthusiasm. Collateral damage, if you will.

    The NVidia microcode wasn't RE'ed because of an overwhelming desire for freedom, it was RE'ed because NVidia did not provide the microcode with a license allowing redistribution. On the other hand I spent time with distro & AMD legal folks to work up a license that *did* meet redistribution requirements, so NVidia HW gets recommended and ours does not.

    Madness.
    bridgman
    AMD Linux
    Last edited by bridgman; 02 May 2015, 02:13 PM.

    Leave a comment:

  • bridgman
    AMD Linux

  • bridgman
    replied
    Originally posted by chithanh View Post
    One is, will your computer run only free software? The answer is no in almost all cases, unless you are running a Lemote Yeelong 8089/8101 with the hard disk removed.
    ... and if you buy into the (IMO mistaken) belief that binary microcode distributed with the hardware is somehow not a problem, even though the same binary microcode distributed other ways (even with controls on updating) is somehow evil.

    Originally posted by chithanh View Post
    Second is, will your computer work with any of the FSF-approved free distros? The answer can be yes if you have an older NVidia card, but no if you have a Radeon card.
    Another circular argument. You're assuming the rules used to create FSF-approved distros are "right" - I'm arguing that they are not only wrong but are sending a message to users that even the FSF is not happy with (although so far they have not yet been sufficiently unhappy to change it).

    Originally posted by chithanh View Post
    If the radeon microcode was opened up, a computer with a Radeon GPU could potentially be used in computers which fit the first criterion. If the microcode was put in flash memory somewhere, it could be used with the free distros per the second criterion.
    The same would be true if the rules used to create FSF-approved distros were updated -- what I would call "fixed"

    And still we have the case where upgradeable microcode in flash is OK but the same microcode in a file is not.

    Leave a comment:

  • SystemCrasher
    Senior Member

  • SystemCrasher
    replied
    In fact I can only see one objection against uCode from disk.

    I can see only one valid reason to dislike uCode on disk more than one in flash. If uCode requires special licensing to redistribute it, it poses trouble for, say, some new distro who can't easily obtain permission to redistribute foreign binary. This sucks and it could be one of valid FSF concerns. Because distro would be crippled in terms of hardware support compared to some more lucky entities who got such permission.

    But as long as vendor permits unlimited distribution of code, it is not anyhow worse than some pre-existing code in ROM/flash/etc. So I guess all this zealotry haves little to do with AMD cards (sure, opensource code could be better, but at least, as long as AMD permits distribution and use of their ucodes. they're not anyhow worse than if they were stored in flash).

    Leave a comment:

Working...
X