Linux GPL and binary module exception clause?
[ Date: Sometime in the near future. ]
[ Scene: Exterior of a Federal courthouse in a large city in the US. Among the cars parked in the lot are several dozen stretch limos, a Saab 9000 Turbo with a penguin bobble-head doll on the dashboard, and a '67 VW van covered with "peace" symbols and sporting a bumper sticker that reads "Code free or die!" ]
[ Scene: Interior of said courthouse. Seated at the plaintiff's table are a gaggle of expensive-looking lawyers in expensive-looking suits. Seated at the defendant's table are Linus T, Alan C, Jeff G, Andrew M, David M, Al V, Richard S, plus a host of other people whose names we might recognize. And one very nervous-looking, pimply-faced young lawyer who looks like he might have graduated from law school sometime last week. ]
[ Lawyer for NVidia: ] "... And in conclusion, Your Honor, we have established that for many years our company sold graphics cards to users of the Windows, Mac, and Linux operating systems, in each case providing a binary graphics driver to make our card work with that OS. Then, without warning, the defendants" [ angrily points his finger at the defendant's table ] "conspired to arrange so that our drivers would no longer work with the Linux OS. We have already demonstrated that, around the same time, our company's revenues began to decline, caused in large part, we believe, by the defendants' actions. We ask for $1 billion in damages."
[ Judge - banging gavel: ] "You've convinced me. I order a summary judgement for the amount requested, plus $2 billion punitive damages."
[ Cut to: Bedroom of a comfortable house in the suburbs. Nighttime. ]
[ Linus - suddenly sits bolt upright in the bed, a horrified expression on his face: ] "AAAAiiiiiiieeeeeeeeaaaaaaarrrrrrgggggghhhhhh!!!!"
[ Wife - shaking Linus' shoulder: ] "Honey, wake up, wake up! I think you're having that horrible nightmare again!"
And that is why binary drivers will always be allowed under Linux.
-p
Where can I get one of those penguin bobble-head dolls?
:))
A question for this thread:
Is the GPL in effect for the kernel so that anybody can enhance the current drivers and add support for any other device? If two companies develop competing products and those products (albeit a few slight differences) perform the same operations using almost the same hardware, do we want one company to use the others driver?
In another sense, does the kernel evolve to reflect this? If the overall driver acts the same minus a few hardware differences, does the kernel source change by abstracting the similarities and allow both companies to write the device specific code? Does it instead say that both cards must have independent source code? Or do we only allow the first driver into the source tree?
There are no evil overtones in this email, nor any disgruntled developer feelings. I am just reading at this thread and asking myself, "Is the overall goal for everyone to get along?"
Dale Whitchurch
In article <1070979148.16262.63.camel@oktoberfest>, Dale Whitchurch <[email protected]> wrote:
| A question for this thread: | | Is the GPL in effect for the kernel so that anybody can enhance the | current drivers and add support for any other device? If two companies | develop competing products and those products (albeit a few slight | differences) perform the same operations using almost the same hardware, | do we want one company to use the others driver?
If company A writes a driver which is not GPL it doesn't concern the
Open Source community. Not even if it's open source but proprietary. Yes, dual license exists, I don't think that changes things here.
If company A writes a GPL driver company B may modify it as long as
they release source.
If company B offered the modified driver for kernel inclusion,
there's a high probability one of the penguins would tell them to fold the changes into the original module and make it dual-purpose (unless there were a LOT of changes).
Company B could decline and ship the GPL driver with their hardware,
source and a binary loadable module included. Given the hassle factor I bet they wouldn't. Nvadia must be really tired of getting every problem related to a tainted kernel.
| In another sense, does the kernel evolve to reflect this? If the | overall driver acts the same minus a few hardware differences, does the | kernel source change by abstracting the similarities and allow both | companies to write the device specific code? Does it instead say that | both cards must have independent source code? Or do we only allow the | first driver into the source tree?
Once GPL'd the choices are clear, it could be separate or added
functionality on a technical basis, no need for one policy to fit all.
| | There are no evil overtones in this email, nor any disgruntled developer | feelings. I am just reading at this thread and asking myself, "Is the | overall goal for everyone to get along?" | | Dale Whitchurch | | - | To unsubscribe from this list: send the line "unsubscribe linux-kernel" in | the body of a message to [email protected] | More majordomo info at http://vger.kernel.org/majordomo-info.html | Please read the FAQ at http://www.tux.org/lkml/ |
Signature · 3 lines
bill davidsen <[email protected]> CTO, TMR Associates, Inc Doing interesting things with little computers since 1979.
Quoted 7 lines
Is the GPL in effect for the kernel so that anybody can enhance the current drivers and add support for any other device? If two companies develop competing products and those products (albeit a few slight differences) perform the same operations using almost the same hardware, do we want one company to use the others driver?
Assuming all the drivers are offered under the GPL, the kernel inclusion
process is a meritocracy. In other words, good code gets added and bad code doesn't.
The GPL provides unrestricted joinder and severance, so if the driver is
GPL and the kernel is GPL, anyone who wants to can join them together under the GPL can do so. Linus would likely do this officially if the driver is reasonable, and anyone who cares enough can do the work needed to get it suitable for inclusion if the original company that offered the driver doesn't want to.
DS
Quoted 7 lines · Dale Whitchurch
A question for this thread:
Is the GPL in effect for the kernel so that anybody can enhance the current drivers and add support for any other device? If two companies develop competing products and those products (albeit a few slight differences) perform the same operations using almost the same hardware, do we want one company to use the others driver?
Sort of. Why not? Especially if the driver is GPL.
Quoted 1 line
In another sense, does the kernel evolve to reflect this?
It evolves to promote standards. If both can use the same driver, then each (if they both contribute) will reduce the development cost of the driver by 1/n, where n is the number of manufacturs of similar boards.
Quoted 4 lines
If the
overall driver acts the same minus a few hardware differences, does the kernel source change by abstracting the similarities and allow both companies to write the device specific code?
Yes. One project that tries to do this for graphics is the GGI. They try to separate the display control registers, timing registers, and data interfacing to allow a mix-n-match the various drivers to come up with a unique combination that would work for a variety of boards.
Quoted 3 lines
Does it instead say that
both cards must have independent source code? Or do we only allow the first driver into the source tree?
This is a combined result. Usually the first one is a specific driver for each device. If the drivers are GPL, somebody else may combine them, or the second manufacturer may take the first driver, make some changes, and release (under GPL) the modified driver. Then both manufacturers (or their engineers) may collaborate to come up with a single driver.
Yes. The multiple mice implementations are an example. So are IDE drivers.
Quoted 3 lines
There are no evil overtones in this email, nor any disgruntled developer feelings. I am just reading at this thread and asking myself, "Is the overall goal for everyone to get along?"
Yup. Everybody contributing. Everybody benifiting.
Quoted 2 lines · Paul Zimmerman
[ Date: Sometime in the near future. ]
[snip]
Quoted 10 lines
[ Cut to: Bedroom of a comfortable house in the suburbs. Nighttime. ]
[ Linus - suddenly sits bolt upright in the bed, a horrified expression on his face: ] "AAAAiiiiiiieeeeeeeeaaaaaaarrrrrrgggggghhhhhh!!!!"
[ Wife - shaking Linus' shoulder: ] "Honey, wake up, wake up! I think you're having that horrible nightmare again!"
And that is why binary drivers will always be allowed under Linux.
If that were the problem, then the kernel would be LGPL, and not GPL. LGPL permits linking (shared libraries), GPL doesn't. To me, it boils down to:
Link with GPL -> result is GPL.
Link with LGPL shared libraries -> result may be anything.
Quoted 13 lines · Jesse Pollard
Quoted 2 lines · Paul Zimmerman
[ Date: Sometime in the near future. ]
[snip]
Quoted 9 lines
[ Cut to: Bedroom of a comfortable house in the suburbs. Nighttime. ]
[ Linus - suddenly sits bolt upright in the bed, a horrified expression on his face: ] "AAAAiiiiiiieeeeeeeeaaaaaaarrrrrrgggggghhhhhh!!!!"
[ Wife - shaking Linus' shoulder: ] "Honey, wake up, wake up! I think you're having that horrible nightmare again!"
Quoted 7 lines
Quoted 1 line
And that is why binary drivers will always be allowed under Linux.
If that were the problem, then the kernel would be LGPL, and not GPL. LGPL permits linking (shared libraries), GPL doesn't. To me, it boils down to:
Link with GPL -> result is GPL.
Link with LGPL shared libraries -> result may be anything.
I don't understand how so much time, effort, and bandwidth can be wasted on Richard Stallman's pet project. The Linux kernel is not about GPL. It's about writing and modifying a high performance operating system that has published source-code.
This allows each and every contributor to demonstrate his or her capabilities in their chosen area of expertise. This gives ideal visibility and demonstrates competence in an area that is replete with charlatans calling themselves programmers.
Already, GPL is preventing this! It started to become evident when Mr. Stallman took credit for the entire operating system and all its utilities. If you don't remember that one, you haven't been awake. He demanded that it be called GNU/Linux, remember? This was the first part of the "divide and conquer" methods that have been used for thousands of years to destroy whole civilizations. This caused a lot of contributors to back off and attempt to isolate their work from the work of others. These factions started to divide up various components including, of all stupid things, header files, and claim that they don't want anybody using their work unless it's under their explicit terms.
I will now cite some information about US copyright law. This information was provided by an "interested third party" who practices intellectual property law. It is not legal advice and is only an opinion. Much of this opinion can be gleaned from "The law of Computer Technology", Raymond, T. Nimmer; Waren, Gorham & Lamont, Inc. ISBN 0-88712-355-4.
The Linux kernel is not an unpublished private work. It has been published for many years. Because it has been published, anybody can use its published components in the manner historically reserved for published works. In other words, you cite your references when you use these published works.
Anybody who commits anything to writing in a manner allowing it to be read back is, in the United States, entitled to the protection of copyright law. One of the protections afforded to the original writer is the ability to control the use of that written work. However, there are limitations set forth by the law. One of the limitations is that the conditions must be "enforceable". Once you publish a work, there are very few things that remain enforceable.
This is because a published work becomes public. That's what publish means. So, all of the information within the kernel becomes public information. Public information is information that is "owned" by everybody (or nobody, depending upon your perspective). The mere act of publishing a work revokes ownership of the information content of that work. The author no longer owns the information because he/she has given it to the public by the act of publishing. This limits what may be considered enforceable. In general, the courts have allowed considerable inclusion of copy-written published works into new works, as long as text has not been inserted "whole cloth".
As previously shown, "#include <kernel.h>" was not a consideration when copyright law was written. In fact, copyright law is poorly suited to software. However, in general, including a header file in software is done to insert information into the work. This information is already public, having come from a published work. It is unlikely that any court would consider restricting this inclusion of information as being enforceable under current law.
If you work for a company and writing software is your job, then you probably have a "standard" header file that your company requires you to use, something like this:
/*
* ################################################################## * # # * # Copyright(C) 1999 - 2003, CatFart Corporation. All rights # * # reserved worldwide. # * # # * # This document contains information proprietary to CatFart # * # Corporation. If this product is acquired by or on behalf of # * # a unit or agency of the United States Government the follow- # * # ing applies: (a) This product was not developed with govern- # * # ment funds; (b) is a trade secret of CatFart Corporation # * # for all purposes of the Freedom of Information Act; and (c) # * # is "Commercial Computer Software". For units of the Depart- # * # ment of Defense (DoD), this software is provided only with # * # "Restricted Rights" as defined in the DoD supplement to the # * # Federal Acquisition Regulations, 52.227-7013(c)(I)(ii). Use, # * # duplication, or disclosure is subject to the restrictions # * # set forth in subdivision (c)(ii) of the Rights in Technical # * # Data and Computer Software clause at 52-227-7013. # * # # * # Use, duplication, or disclosure of this proprietary document # * # without the express permission of CatFart Corporation is # * # prohibited. # * # # * ##################################################################
*/
This shows that the software contained within is highly-restricted. This restriction remains because this document is unpublished. It is an unpublished, private work of CatFart. Therefore any use or disclosure is enforceable. To publish this information, requires that I obtain the permission of CatFart. Just to be safe, I need to state, "Reproduced by permission of CatFart Corporation", any time I reference the contents of this document. Wanna register CatFart.com?
Once I publish the contents of this document, the statement in the header is moot (invalid). You can't have it both ways. You either have an unpublished private document that may contain secret information or you have a public document that contains no secrets whatsoever.
Note that I just gave away "CatFart.com" by publishing this information. That's what "publish" means.
Given that the publishing of a work gives away the information in that work, one may wonder why anybody publishes anything. But that's a topic for a whole new discussion, hopefully, not in the Linux-Kernel list.
Developers need to back off and stop getting hot-under-the-collar about things that are quite beyond their control. Neither SCO, IBM, nor Stallman can take credit for your work although they may try. You need to make it easy for others to use your work so that it remains visible. Your own perspective should grow with the proliferation of Linux, not collapse into a tunnel- vision of; "They ain't gonna use my stuff...". There is a lot of truth to the adage, "Publish or perish.". You chose to contribute to a published work, great! Now, don't pretend that you own it. You did, after all, contribute it to the greater good of all, hoping so put M$ and other pretenders to shame.
This EXPORT_SYMBOL_GPL stuff is absurd. Anybody who knows how to link object files together can bypass any attempted "protections" altogether. There is entirely too much of this crap going on. Surely, one needs to know if there is some unpublished binary lurking in the kernel that could be screwing up the works, but beyond that playing with symbols simply makes the developers look like fools.
Cheers,
Dick Johnson
Penguin : Linux version 2.4.22 on an i686 machine (797.90 BogoMips).
Note 96.31% of all statistics are fiction.
Jesse,
Linking to become one with the vmlinux (zen thing) or not able to achieve a modular state, you are toast. Loading a module is not linking. Now people claim that /proc/kcore is where the dirty work happens.
Is "/proc/kcore" real?
What makes it real? Who makes it real?
If you, the user of the binary module, execute:
cat /proc/kcore > /kcore.file
Who combined the works?
It was not the author(s), it was the effective enduser.
Cheers,
Andre Hedrick
LAD Storage Consulting Group
Quoted 26 lines · Jesse Pollard
Quoted 2 lines · Paul Zimmerman
[ Date: Sometime in the near future. ]
[snip]
Quoted 10 lines
[ Cut to: Bedroom of a comfortable house in the suburbs. Nighttime. ]
[ Linus - suddenly sits bolt upright in the bed, a horrified expression on his face: ] "AAAAiiiiiiieeeeeeeeaaaaaaarrrrrrgggggghhhhhh!!!!"
[ Wife - shaking Linus' shoulder: ] "Honey, wake up, wake up! I think you're having that horrible nightmare again!"
And that is why binary drivers will always be allowed under Linux.
If that were the problem, then the kernel would be LGPL, and not GPL. LGPL permits linking (shared libraries), GPL doesn't. To me, it boils down to:
Link with GPL -> result is GPL.
Link with LGPL shared libraries -> result may be anything.
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [email protected]
More majordomo info at http://vger.kernel.org/majordomo-info.html
Please read the FAQ at http://www.tux.org/lkml/
Quoted 5 lines · Andre Hedrick
Jesse,
Linking to become one with the vmlinux (zen thing) or not able to achieve a modular state, you are toast. Loading a module is not linking. Now people claim that
/proc/kcoreis where the dirty work happens.
So loading a module doesn't include it in any tables, and hence modifies them? It doesn't provide or accept relocation information to alter address references?
It doesn't need to be invoked by the kernel?
Looks nonfunctional to me.
Quoted 9 lines
Is "
/proc/kcore" real?What makes it real? Who makes it real?
If you, the user of the binary module, execute:
cat
/proc/kcore> /kcore.fileWho combined the works?
Not relevent. Everything is done in user mode using defined user mode interfaces for kernel/user mode separation.