Binary-only modules not to be banned

Yesterday, it looked as if the kernel developers were heading toward banning the loading of binary-only modules. Since then, Linus has come out against the change, and, after some discussion, Greg Kroah-Hartman has withdrawn the patch, saying "Even if we change the kernel this way, it prevents valid and legal usages of the kernel." The frustration which drove this discussion remains, however.

to post comments

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 14:33 UTC (Thu) by pbardet (guest, #22762) [Link] (4 responses)

Am I glad to still have the freedom of adding whatever I want to my kernel !
For a little while, I was worried that FOSS police would replace MS police on my computer.

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 15:00 UTC (Thu) by arny (guest, #7185) [Link] (3 responses)

According to the GPL, you would still be allowed to add whatever you want to your kernel.

But you won't be allowed to _give away_ non-gpl kernel derived components. In other words, companies won't be allowed to give you their proprietary drivers.

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 16:30 UTC (Thu) by khim (subscriber, #9252) [Link] (1 responses)

In other words, companies won't be allowed to give you their proprietary drivers.

That's why they generally don't do this. Canonical is at fault, not nVidia...

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 21:58 UTC (Thu) by drag (guest, #31333) [Link]

Well Nvidia is still borked.

They themselves are in this legally dubious 'derived works' area with their drivers that they do ship. They still do binary-only drivers, the comaptability shim is not a loophole in the GPL.

If they supported Linux properly they would either open source the drivers or help Linux developers out with documentation and other assistance so that the people working on open source nvidia drivers don't have to waste all their time hunting down bugs in the hardware.

I don't have a problem with people using Nvidia out of nessicity.. you have to do what you have to do. Even if Nvidia is within the realm of legality it still entirely means that they are at fault for not supporting linux properly. (in other words, just because it can be twisted into being legal doesn't mean it's right)

Binary-only modules not to be banned - this time

Posted Dec 18, 2006 19:33 UTC (Mon) by CowboyJack (guest, #42308) [Link]

I read the discussion on this,
and disagree with your analysis.

The "patch":
looks like a DMCA technical measure,
acts like a DMCA technical measure,
smells like a DMCA technical measure,
and tastes like a DMCA technical measure.
Therefore (thanks to my College Logic course):
the Patch *is* a DMCA technical measure.

Hence a mere user could go to jail for disabling the
technical measure. Good bye "Free" and "Open Source" software!

Thank God, rationality won out.

However, I understand the reasoning behind the desire to do this,
but there needs to be a better answer. Perhaps, an
opt-in "phone home" message that notifies copyright owners of
binary modules that violate the copyrights of others. This
would give a report of damage which authors could use to shut down violators. I'd turn this on in all my PCs if it was there.
I believe in copyright - as broken as it it.

G'day,
Too-long-reply Brian

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 14:34 UTC (Thu) by coriordan (guest, #7544) [Link] (1 responses)

I think it's an exageration to say that Linus came out against it. When Linus is against something, there's *lots* of ASCII _markup_ and mentions of NEVER NEVER NEVER and "fundamental" and "so lets not even discuss this".

Given that style, his actual comments seem like he's open to debate.

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 16:01 UTC (Thu) by madscientist (subscriber, #16861) [Link]

I think it's an exaggeration to say he's open to debate :)

What he said was that he absolutely would not merge it "from the top down"; that is he wouldn't put it into his tree and in that way push it out to all the other distros, etc. He doesn't seem open to any debate on this subject.

He did say that if all or most of the vendors were to add it to their trees and support it as a "standard" add-on patch, he'd bow to the will of the majority and merge it, "from the bottom up", into his tree. But that he'd still think it was a dumb idea :)

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 15:42 UTC (Thu) by rvfh (guest, #31018) [Link] (10 responses)

Pfeww! For a while I thought my freedom to use WHATEVER was gone. No more 3D on my desktop would really bother me! And no, I haven't got the skills to work on 'nouveau' (but I'll switch with great joy when they get something kinda working).

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 15:48 UTC (Thu) by alexbk (subscriber, #37839) [Link] (9 responses)

Surely you'd be able to revert the change, since all the source is out there, or persuade your vendor to do it, so your freedom was never gone.

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 21:36 UTC (Thu) by huaz (guest, #10168) [Link] (8 responses)

So you are saying only people who can rebuild their own kernels would get the freedom back.

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 22:02 UTC (Thu) by Arker (guest, #14205) [Link] (7 responses)

In a limited sense, that's true, and in that sense, it's true regardless. Freedom of the Press belongs to those that own one.

Of course you don't have to compile it yourself to take advantage of the fact that each distributor compile's it with the patches they believe are most valuable for their customers, just as you don't have to own your own press to benefit from the freedom of those who do to publish without censorship.

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 22:37 UTC (Thu) by drag (guest, #31333) [Link] (6 responses)

I don't remember anybody ever going down with their ship in battle saying:

"Give me convenience or give me death!"

Although I dont' beleive they should use technical measures against binary-only drivers, especially since it's obviously going to cause problems for legal use.

But remember that freedom isn't free, and there is no reason why it has to be convient also.

I wouldn't see anything wrong with having a patch that outputed a explanation in dmesg about how if they got this driver pre-compiled from somebody else that this is indeed a violation of the linux developer's copyright and they would appreaciate it if you tell them to stop doing that.

Binary-only modules not to be banned - this time

Posted Dec 15, 2006 3:14 UTC (Fri) by Arker (guest, #14205) [Link] (5 responses)

I'm not sure how that would cause problems for any legal use. Any user or distributor is free to strip that code out, after all. It would make it harder for people to run tainted kernels without being aware of the problem, however.

Binary-only modules not to be banned - this time

Posted Dec 15, 2006 3:32 UTC (Fri) by drag (guest, #31333) [Link] (4 responses)

Well the awareness is the issue for me also.

A lot of people are going to run drivers that they don't understand that nobody can fix them or troubleshoot them if something goes wrong.

I am thinking of new users and such. Somebody new to Linux and isn't very technical could end up buying a machine or obtaining a peice of hardware that has a propriatory driver and they don't understand the problems you can run into with something like that.

Say somebody buys a Redhat server for home or something like that and something goes wrong as he is getting file system corruption.

He goes to Redhat and it turns out for the SATA RAID he is stuck using a propriatory kernel module.

How does something like that get resolved? Redhat can't fix it for him even if he has a contract with Redhat.

If that person had bought hardware whose maker supported Linux properly then it could get resolved and he wouldn't have to worry about Redhat corrupting his data anymore. Or if Redhat didn't work out he could of gone to somebody else and had it fixed.

Surely if he known ahead of time about it he would of gotten a real sata raid controller or used MD, or gone with a different vendor.

But how can this sort of thing be avoided?

Binary modules for hardware

Posted Dec 15, 2006 7:40 UTC (Fri) by man_ls (guest, #15091) [Link] (3 responses)

Having a piece of hardware which can only be controlled by a proprietary module is not much different than having a piece of hardware which has no Linux support at all. Does that situation have a solution?

In fact it is a very limited subset of supported hardware, and in fact it has been argued that proprietary modules cannot be called "Linux support" properly. Finally, from what I have heard I don't think that Red Hat would be very happy to let you use a proprietary module today.

So you would just say: "Sorry, that piece of hardware is not supported; buy another SATA RAID kit" and be done with it.

Binary modules for hardware

Posted Dec 15, 2006 9:38 UTC (Fri) by Arker (guest, #14205) [Link] (2 responses)

The trouble is that, although it's an obvious lie to anyone that understands the situation, people *do* sell unsupported software as supported. Something like this could help to discourage that. Right now, you can go out and buy a machine using new ATI or Nvidia video cards that cannot be supported under linux, in machines sold specifically for linux use. You and I may know enough to check the specs and notice the lie, but the more new users come into the community the more of us are vulnerable to such tactics. Anything that makes that line between truly supportable under linux and might work with linux if you corrupt your kernel with an opaque blob clearer is a good thing IMOP.

Binary modules for hardware

Posted Dec 15, 2006 11:06 UTC (Fri) by drag (guest, #31333) [Link] (1 responses)

Ya.

Plus the goal is to avoid having people in that situation in the first place.

It's difficult as you will have something like a Adaptic 'fakeraid' controller that actually advertises Linux support and you'll have people in the Windows world that swear up and down how wonderfull it is.

So a carefull person, even if they are relatively knowledgable about most things but a bit naive in Linux will go out and buy it thinking they made a wise choice, which under normal circumstances they did.

Maybe I am being overly dramatic becuase I dislike the idea of having a open source os depend on closed source software for even basic functionality, but it can be a difficult problem it seems to me.

Binary modules for hardware

Posted Dec 15, 2006 12:10 UTC (Fri) by Arker (guest, #14205) [Link]

I've seen this happen way too often, and I've seen too many people come away thinking Linux is crappy because of it. It's much better to just be able to say 'that device isn't supported' than to have a situation where you *might* be able to make it work, but only if you stay within the narrow confines that the manufacturer produces a blob for, and only at the cost of throwing nearly all the natural advantages of using free software down the drain. And it's particularly horrid when you have to explain to a customer that a particular piece of hardware they already paid good money for really isn't supported under linux, despite the manufacturers claims to the contrary.

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 15:48 UTC (Thu) by Richard_J_Neill (subscriber, #23093) [Link] (16 responses)

I think, in the end, we'd be better off without the proprietary drivers. I'd particularly like to see support for the nvidia/ati drivers withdrawn, because it would then provide a real motivation and market-advantage to any company that provides open ones.

Personally, I use the nvidia driver because the nv driver can't do dual-head, but I'd buy any new (reasonably cheap) graphics card that can give me dual-head DVI at 3200x1200 with a free driver.

Linus has a good point in that, if the kernel becomes closed-driver hostile, it will push some companies into a "user-space shell game", and make others withdraw their code entirely. BUT, if we keep permitting closed drivers, then more and more new products will ship with binary drivers.

Ultimately, Stallman is right: freedom is a binary choice - you either have it, or you don't - and there is no long-term viable half-measure.

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 16:48 UTC (Thu) by dlang (guest, #313) [Link] (14 responses)

there isn't any support for these drivers in the kernel (other then the ability to load modules) so I don't understand what support you think should be removed.

also, these video drivers (while extremely painful to deal with) are also the best example of drivers that are not derived from the kernel, they are ported to the kernel interfaces but were written for that other OS.

now I recognise that it is possible to have something derived from multiple sources, and as a result it is possible that video drivers ported from elsewhere could be also derived from the kernel, but especially in the case of nvidia where they have the glue layer to provide their own API to their real driver that driver itself is not derived from the kernel.

and if something isn't derived from the kernel copyright law cannot apply to it, and as such the GPL (v2 or v3) cannot prohibit it.

for the record, I do not purchase any nvidia cards for my systems, I am able to stick with the older ATI cards that do have open drivers. Even though I strongly suspect that the closed drivers are legal, I just don't want the compatability issues with binary drivers on my systems.

if something isn't derived from the kernel ...

Posted Dec 14, 2006 17:33 UTC (Thu) by edmundo (guest, #616) [Link] (1 responses)

> if something isn't derived from the kernel copyright law cannot apply to it, and as such the GPL (v2 or v3) cannot prohibit it

Not necessarily. If the binary driver is useless without the Linux kernel, then someone who distributes the binary driver might be considered to be in effect distributing the derived work that consists of the driver plus the kernel.

If this were not the case, it would be possible to circumvent the GPL in other cases, too. For example, if you wanted to make a non-free compiler based on GCC you could get someone else to turn GCC into a GPL library and distribute that library while you distribute a binary that links against the library while claiming that the GPL does not apply to you because you are not personally distributing the GPL code.

if something isn't derived from the kernel ...

Posted Dec 14, 2006 21:47 UTC (Thu) by Soruk (guest, #2722) [Link]

Not quite.

Dynamic linking with GPL code like this would violate the GPL if the code trying to link to it wasn't GPLed (or GPL-compatibly licenced, e.g. MIT).

Note the licence for Cygwin is GPL, cygwin1.dll itself is also GPL, NOT LGPL, which has resulted in Cygnus/RedHat adding the exception clause that linking with software licensed by any OSI-approved licence is permitted. They would not have to do that if what you say is true.

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 18:35 UTC (Thu) by Richard_J_Neill (subscriber, #23093) [Link] (10 responses)

> there isn't any support for these drivers in the kernel (other then the
> ability to load modules) so I don't understand what support you think
> should be removed.

Sorry, that wasn't very clear. I'm not making a technical suggestion but a legal one: Linus has always said that he considers binary-drivers to be a grey area, but has tacitly tolerated them. However, I think that the driver-authors (nvidia et al) have created a derived work of the kernel, and are therefore *probably* in breach of the GPL. Ideally, we'd see a GPL lawsuit to demand that nvidia etc provide the source to match the drivers which they have already released.

BTW, re older ATI cards with free drivers, can any of them still be purchased? I thought that it was impossible to buy a currently-sold dual-head DVI card (even without 3D-accel) which had open drivers.

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 21:23 UTC (Thu) by dlang (guest, #313) [Link] (8 responses)

The point that I am makeing is that just becouse the nvidia linux drivers are useless without the linux kernel doesn't nessasarily make them derivitive works of the kernel. In the same way a driver written for windows isn't a derivitive work of the microsoft kernel.

the core things in the nvidia drivers (what everyone cares about) are ports of their windows drivers, not modifications of anything that exists (or existed) in the linux kernel.

just becouse something is made into a module that can be loaded into linux it doesn't mean that it was derived from linux

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 22:24 UTC (Thu) by drag (guest, #31333) [Link] (7 responses)

What you guys are in effect saying is that it is alright for me to link a closed source program to a GPL library as long as it was closed source before I linked it?

It doesn't make sense.

So say Gstreamer was GPL licensed (not LGPL like it realy is).

So I can write a closed source audio application for Linux that then would link against Gstreamer just as long as provided a 'shim' of open source software as a buffer between my closed source application and the GPL library.

If it's true then what is the point of having LGPL at all? People can just link to and use GPL'd libraries without any fear because they can just say "it's not derived work'?

They could easily work around any licensing restrictions in any gpl software. All they have to do is in the download they have the installer automate compiling a few lines of code and tell people that "oh most of the code in this app is also in the my windows version so it's ok'

That is EXACTLY what is going on here. It doesn't make any sense what you guys are saying and Linus is just wrong.

It just doesn't make any sense.

on a side note:
Now if Linus and friends want to allow linking of binary drivers under certain circumstances, then that is fine. It's obvious that they don't realy mind that much.

also I absolutely beleive it is the right of anybody to shovel whatever code they want into their own software, closed source or not. If you want to dump the entire Win98 code base into the Linux kernel or anything else then you won't have any agrument with me at all.

Linus thinks it's a grey area and he's not wrong.

Posted Dec 15, 2006 3:10 UTC (Fri) by xoddam (subscriber, #2322) [Link] (6 responses)

> If it's true then what is the point of having LGPL at all? People
> can just link to and use GPL'd libraries without any fear because
> they can just say "it's not derived work'?

The combined work which is finally executed is *clearly* a derived work
of both the library and the other code which links to it. Preparation
and use of that derived work by the end user at execution time is
permitted by the GPL (version 2 at least). But anyone attempting to
distribute the two pieces of code "as part of a whole" -- all potential
integrators, not necessarily including the author of the client code --
is required to distribute both under the GPL.

On a superficial reading, the opinion of the FSF *seems* to be that if
you write code which calls an API which is provided by a GPL library,
then the GPL must cover your code.

http://www.gnu.org/licenses/why-not-lgpl.html

Yet this is not explicit in the letter of the GPL; as I read it the
licence has nothing to say if the two components are not distributed
together. Just because the LGPL explicitly permits this usage (which is
obviously important for integrators) doesn't mean the GPL unambiguously
prohibits it.

It is not clear to me that there is any way to argue this point except to
claim that the API-using code is a derived work of the API
implementation. This does not gel with the standard legal interpretation
of "derived work" (otherwise you'd need a copyright licence for every
interface you use), hence Linus' insistence that it is a grey area.

The FSF does not claim anywhere to have made this argument. It may be
that the "success story" (that they persuaded someone to GPL a program
because they wanted to use the GPL'd readline library) was based only on
the argument that the third-party code could not legally be distributed
with the GPL library unless it was also under the GPL, without making any
attempt to claim that the one was a derivative of the other.

> They could easily work around any licensing restrictions in any gpl
> software. All they have to do is in the download they have the
> installer automate compiling a few lines of code and tell people
> that "oh most of the code in this app is also in the my windows
> version so it's ok'

This issue has never been tested in court. When the GPL has been
*clearly* violated, it has clearly won, but this is not a clear
violation.

> That is EXACTLY what is going on here. It doesn't make any
> sense what you guys are saying and Linus is just wrong.

Linus is wrong to say it's a grey area and he likes it that way?
It doesn't make sense what you are saying, I think you are just wrong.

Linus thinks it's a grey area and he's not wrong.

Posted Dec 15, 2006 3:55 UTC (Fri) by drag (guest, #31333) [Link] (5 responses)

Legally gray isn't nessicarially something that is nice to be gray.

I think it's much nicer to know what exactly your allowed and not allowed to do. This makes things much easier for everybody.

But it's a reality you can't avoid. So gray it is.

But we aren't talking about some thing like Wine or POSIX interface were you can program a application that can link up and run on a veriaty of operating systems.

This isn't a generic standardized interface. Those video drivers programmers are sitting there with the linux kernel source code and writing code specificly designed to work with the kernel and nothing else. Sure a lot of that binary stuff is going to be generic for Windows and whatnot, but I expect a lot of it isn't.

On the flip side I don't think it's nessicary that Nvidia has to open source their drivers either. The kernel developers set out specific exceptions for people to use portions of the kernel for 'public interfaces' vs 'gpl-only' code.

Essentially the kernel isn't licensed GPL. It's licensed similarly to how Sun's Java implimentation is going to be were you have 'GPLv2 + Classpath exceptions'.

By only using these 'public interfaces' then nvidia isn't violating anything, I suppose.

Like I said if the kernel developers want to allow binary drivers to link against the kernel then that's their choice.

(If Linus beleives the 'gray area' thing, then what is the point of going 'this and this' is ok to use, but that isn't?)

But I have to disagree with you on the whole derived thing. I think that the distinction between 'dynamic' linking and 'static' linking is irrelevent and your using their code in your software when you use their libraries in that way.

Linus thinks it's a grey area and he's not wrong.

Posted Dec 15, 2006 4:51 UTC (Fri) by xoddam (subscriber, #2322) [Link] (3 responses)

> I think it's much nicer to know what exactly your allowed
> and not allowed to do. This makes things much easier for
> everybody.

It's easier to do something if nobody knows that you're
not allowed to do it, than if everyone agrees you're not.

> Those video drivers programmers are sitting there with the
> linux kernel source code and writing code specificly designed
> to work with the kernel and nothing else. Sure a lot of that
> binary stuff is going to be generic for Windows and whatnot,
> but I expect a lot of it isn't.

Well if you want to make your code work with something, you have
to design it that way. I do not believe that doing so makes your
code a derived work of what you're linking to. If the particular
driver you're talking about is a derived work based on an informed
abstraction-filtration-comparison test, I'll happily concede that
the authors are violating the GPL by distributing them. Nothing
else will prove it; using an API which is implemented by GPL'd
code is not sufficient.

> But I have to disagree with you on the whole derived thing. I
> think that the distinction between 'dynamic' linking and 'static'
> linking is irrelevant

I didn't distinguish between two kinds of linking; the technical
measures involved are not relevant to the copyright law as far as
I'm aware. However something which does not incorporate the
implementation of an API it uses is not derived from the implementation.
The act of linking obviously does prepare a derived work of the
things you link together, and so might other acts like simply
throwing the two things together on a CD (Greg KH certainly thinks
so). In certain jurisdictions you need permission (usually implicit)
even to prepare a derived work. The GPL version 2 grants this
permission explicitly, and places conditions only on distribution.
If the licence of the non-GPL part is incompatible with the GPL,
distribution of the combined work is prohibited altogether.

> and you're using their code in your software
> when you use their libraries in that way.

Interfaces are pretty explicit boundary lines. I do not believe
that a user of an interface is a derived work of the implementation.

The why-not-LGPL page and your discussion above both seem to imply
that if the interface has a unique implementation, then any user
*is* a derived work, while interfaces with multiple implementations
are somehow exempt from this rule.

I think if anyone tried to claim this explicitly, their argument
would be torn to shreds on short notice.

Most interfaces start out with a unique implementation and gain
others as a result of perfectly legal interoperability efforts.
The Linux kernel module interface is huge, changes all the time,
and has no interoperable competitors, but it isn't privileged
on copyright grounds.

What *is* your argument exactly?

Linus thinks it's a grey area and he's not wrong.

Posted Dec 15, 2006 5:07 UTC (Fri) by drag (guest, #31333) [Link] (2 responses)

""What *is* your argument exactly?""

That if it's illegal to link against GPL'd software and not follow the terms of that license then adding open source shims and playing games with what is derived vs what is not derived is just silly pointless stuff, it's still a violation of the copyright license.

But I guess you beleive that it's fine to link against GPL software and not follow the terms of the GPL software.

So we are at a impasse.

Linus thinks it's a grey area and he's not wrong.

Posted Dec 15, 2006 7:18 UTC (Fri) by xoddam (subscriber, #2322) [Link] (1 responses)

> That if
> [it's illegal to link against GPL'd software and not follow the
> terms of that license]
> then
> [adding open source shims and playing games with what is derived
> vs what is not derived is just silly pointless stuff, it's still
> a violation of the copyright license.]

It's a valid syllogism, but I do think that the premise
<it's illegal to link...> is false.

> But I guess you beleive that it's fine to link against GPL software
> and not follow the terms of the GPL software.

Erm ... I believe that it doesn't violate the GPL to develop
something which links with GPL software, and that it still
doesn't violate the *letter* of the GPL to release what you
have developed under the licence of your choice, including a
GPL-incompatible licence, providing you don't distribute
something derived from the GPL software.

There's an argument to be made that it violates the *spirit* of
the GPL. Personally I don't think it does, because the stated
purpose of the GPL is to preserve the freedom of the end-user with
respect to the licenced work. The restrictions on distributors
are to ensure that freedom is preserved for end users, and that
GPL'd code is not swallowed up within a proprietary architecture
without which it is useless. It is up to users to decide what
use to make of this freedom (and how much of it they are willing
to trade for convenience). ATI, nVidia and their customers are
all users in this context.

> So we are at a impasse.

The FSF claims that GPL-licenced libraries are not available for
non-free software to use. This case is one of the weakest I've seen
them make and I'm unconvinced that it has a basis in law (besides
the barrier to entry imposed by integrators and distributors, who are
barred from distributing combined works with incompatible licences).
Normally I'm a strong FSF supporter, but I'm just not convinced in
this instance.

The real reason I weigh in on the subject at all is that people seem
to be adducing reasons for the illegality of binary kernel modules
which haven't actually been made by the people who claim that the
binary modules violate their copyright. The Kororaa shutdown came
down to the fact that their live CD included a kernel image together
with ready-made GPL-incompatible driver modules:

http://kororaa.org/index.php?entry=entry060521-200059

Various people have said words to the effect that the mere existence
of a GPL-incompatible module violates kernel copyright. None of
them has made a convincing case for this, though. The most coherent
argument I've heard involved the complexity of certain kernel header
macros, here on lwn.net when discussing Kororaa:

http://lwn.net/Articles/184885/
http://lwn.net/Articles/185169/

I'm not entirely convinced, and I've never seen any of the actual
copyright holders or lawyers try to argue on this basis.

Linus thinks it's a grey area and he's not wrong.

Posted Dec 15, 2006 11:15 UTC (Fri) by drag (guest, #31333) [Link]

Don't worry. you've given me good things to think about.

It's also worth noting that Kororaa is still distributing it's software, but it is now using open source drivers to do it. They basicly said the same thing you did which is that even though it's gray it's against the spirit and general wishes of the kernel team so they decided to stop shipping binary driveres.

I think that in situations like this intent means a lot. I think that if the copyright owner of the GPL'd work makes it very obvious when questioned that they do not allow linking of propriatory software then it's going to have a different effect then if the author was more permissive.

But I am so far from a lawyer that it's not even funny. So what do I know?
(not a whole lot)

Linus thinks it's a grey area and he's not wrong.

Posted Jan 2, 2007 18:29 UTC (Tue) by arafel (subscriber, #18557) [Link]

> Those video drivers programmers are sitting there with the linux kernel
> source code and writing code specificly designed to work with the
> kernel and nothing else.

For the main driver you don't know that, unless you've seen the source. I doubt you have.

The general argument is that if you have an abstraction layer which provides a common interface for (say) making a memory region accessible for the driver to read/write, hook an interrupt, down/up mutexes etc. you've covered more or less everything needed by the driver.

Note that these are all generic operations; I don't see any possible way to claim they're specific to GPL code and that anything using them is a derived work.

Any Linux specific bits - registering a chardev, for example - can be handled by a pure GPL "wrapper". This wouldn't affect the driver code, because it shouldn't care what's using it - it shouldn't have been written with any particular wrapper in mind.

What about two PCI cards?

Posted Dec 15, 2006 12:54 UTC (Fri) by grantingram (guest, #18390) [Link]

I thought that it was impossible to buy a currently-sold dual-head DVI card (even without 3D-accel) which had open drivers.

I've never tried this but what about two Radeon 9250 PCI cards? That way you would get two DVI outputs and open 3D acceleration!

In the UK you can buy these for about £30 now.

Of course I haven't actually done this and am a total raving incompetent when it comes to hardware - so you'd be insane to take my advice without talking to someone who actually has a clue....

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 22:52 UTC (Thu) by davej (subscriber, #354) [Link]

>also, these video drivers (while extremely painful to deal with) are also
>the best example of drivers that are not derived from the kernel, they are
>ported to the kernel interfaces but were written for that other OS.

I suggest you look through the ATI driver a little more closely.
In particular note the agpgart parts that it statically links to its proprietary code with some nice glue that when diffed against the kernel agp code looks something like ..

+#ifdef STANDALONE
MODULE_LICENSE("GPL");
+#endif

not only is it completely clear that it's a derived work, it's pretty clear to anyone with half a brain that it's subverting the GPL license on the code it links to.

Binary-only modules not to be banned - this time

Posted Dec 15, 2006 6:02 UTC (Fri) by tjw.org (guest, #20716) [Link]

Personally, I use the nvidia driver because the nv driver can't do dual-head, but I'd buy any new (reasonably cheap) graphics card that can give me dual-head DVI at 3200x1200 with a free driver.

I'm looking for the same solution myself and the new Intel offering (G965 chipset) looks pretty good.

This is an on-board only graphics chip and you won't find a motherboard that has dual-dvi, but you can use what's called an ADD2 card in the PCI-E slot to use dual-dvi with the on-board graphics chip:

http://www.newegg.com/Product/Product.asp?Item=N82E16815133002

The only things holding me back are the fact that this chipset and boards that use it require a bleeding-edge kernel and Xorg/glx to work and that I have found very little in the way of testimonials of people using this configuration.

However, it's great to see Intel's effort for open Linux drivers. They even appear to be funding Xorg development for their drivers:

http://www.intellinuxgraphics.com/team.html

David Woodhouse calling in the lawyers - on who?

Posted Dec 14, 2006 17:12 UTC (Thu) by Randakar (guest, #27808) [Link] (1 responses)

The _really_ interesting comment to fall out of the tree after this vigourous shake is the one David Woodhouse made regarding making ready to sue some company regarding their binary driver.

Not sure what such a counter-offensive would accomplish (making companies more skittish about coding drivers for Linux, I'd imagine) but it's sure going to be interesting to watch ..

As for the original topic of discussion; Personally I absolutely would love to have fully GPL'ed drivers for everything. Merely having this discussion might help put a bit more pressure on driver manufacturers. I'm not holding my breath, though. As usual Linus is being the cool and clear headed person again in coming out against the change.

Still it's a shame; somehow there has to be a way to nudge/push these companies into opening up their drivers - I just fail to see it.

David Woodhouse calling in the lawyers - on who?

Posted Dec 14, 2006 22:49 UTC (Thu) by drag (guest, #31333) [Link]

As a Linux user don't buy their crap. As a kernel developer you just make a big fuss, write few letters, knock a few heads.

Whenever people ask about what hardware they are going to buy, even if it's for Windows right now.. tell them that if they ever want to run Linux in the future for any reason that they shouldn't buy it.

With the exception of fast 3d acceleration there is no reason why you have to have closed source drivers anywere on your system. You can get RAID cards, Sata controllers, Wifi controllers, motherboards, data aquasition cards, video encoders, HDTV capture cards, etc etc.

You name it, you can get it open source drivers for it, for a normal PC. You just have to be smart about the hardware you buy.

And _yes_ I've been burned by buying the wrong hardware, sometimes it's not easy. But usually it doesn't require much work.

It just realy realy realy sucks that the whole desktop is dependant on one hostile hardware vendor: Nvidia.

Despite all the work people from Novell are doing, and Portland project, and Freedesktop.org, X.org, kernel developers, etc etc. All of it can be crippled just by them. It's a bad place to be.

Binary-only drivers are the bad pixels of the 21st century.

Posted Dec 14, 2006 20:00 UTC (Thu) by dmarti (subscriber, #11625) [Link]

Remember "no bad pixels" clauses in RFQs and purchase orders? People who buy Linux hardware for work could do something similar for a "no tainted drivers" clause. Please let me know if you would attach such a term to any purchase orders or RFQs you send out, and I'll get one drafted. Once these purchase orders start coming in, then Sales at the server vendor bugs Engineering at the server vendor who bugs Sales at the component vendor who bugs Legal at the component vendor who fixes the problem. The shortest solution to any paperwork problem comes when the problem stands between a sales person and a sale.

Binary-only modules not to be banned - this time

Posted Dec 14, 2006 22:01 UTC (Thu) by mjthayer (guest, #39183) [Link] (9 responses)

See this page - https://launchpad.net/bounties/free-nvidia-drivers - for
a small experiment.

Binary-only modules not to be banned - this time

Posted Dec 15, 2006 0:18 UTC (Fri) by lamikr (subscriber, #2289) [Link] (8 responses)

What was the status for Intel's free video 3D drivers that were announced
some time ago. Are they really gpl, I remember reading something about big binary blob? And even more important, is it possible to buy a core 2 duo laptop with those Intel graphics cards?

Binary-only modules not to be banned - this time

Posted Dec 15, 2006 3:10 UTC (Fri) by Arker (guest, #14205) [Link] (6 responses)

It's a mix of GPL and MIT (GPL compatible) licensing, and yes, there are several laptops for sale using that video set, and several with the coreduo CPU and intel video:

- the Apple MacBook
- the Sony Vaio BX64D/66D, TX85D, TXN17
- the Dell Latitude D420
- the Sharp Actius M4000

and, I'm sure, many more. I didn't want to spend too much time hunting them down for you, I'm sure you can figure out how to use google too.

The key is just to search, and make sure you know exactly what you're buying. Different laptops are often given the same or confusingly similar names.

Emperorlinux.com, linuxlaptops.org, and linuxquestions.org may be useful in the search.

Binary-only modules not to be banned - this time

Posted Dec 15, 2006 3:24 UTC (Fri) by drag (guest, #31333) [Link] (3 responses)

Goto system76.com and get one for intel onboard chipset. They are pretty good.

I am not sure about the wifi if it's one that requires the intel 'regulator' binary blob or not. It would be nice to avoid that.

A couple notes on the Intel stuff to keep in mind:

There was a small binary blob for the Intel driver, but it doesn't affect the operation of the driver if you use it. Basicly it's for things like 'Macrovision' content protection and such, you can safely ignore it and I don't think anybody is shipping it.

Also another downside is that Intel only makes the documentation aviable under NDA, which limits what people can do to fix problems.

This is unfortunate, but it is still much better then what we get from either ATI or Nvidia.

Binary-only modules not to be banned - this time

Posted Dec 15, 2006 10:59 UTC (Fri) by Arker (guest, #14205) [Link] (2 responses)

Nice tip, thanks.

Do you know who actually manufactures their laptops?

Binary-only modules not to be banned - this time

Posted Dec 15, 2006 11:25 UTC (Fri) by drag (guest, #31333) [Link] (1 responses)

I would hazard a guess that Quantas would, or maybe Asus.

I haven't owned one, but they are pretty cool abotu things in interviews that I've heard. Even though they ship nvidia binaries I don't think it can be held against them. Consumer demand and all that.

Even they go so far as to try to help out developing open source drivers for their unsupported hardware, such as the onboard camera.

Their support seems good. A year. Then they have a 'Ubuntu driver' which they maintain that has the drivers that aren't aviable in the default Ubuntu kernel that their hardware may need. They seem to just stick it all in one big deb file you can install.

On the TLLTS they interviewed one of the developers on Show #163. So you may want to listen to that first and make your own opinion.

On show #171 one of the show hosts ended up actually buying one of the laptops. I am listenning to that right now so I don't know how it went yet.

http://tllts.org/

they even have ogg feeds. (makes me happy)

Binary-only modules not to be banned - this time

Posted Dec 15, 2006 12:00 UTC (Fri) by drag (guest, #31333) [Link]

Actually I was googling around for it and appears that System76 is ASUS-based laptop.

Which sounds good for me. I have a couple Asus things and I've liked them so far.

Keep in mind that there are 'ODM's vs 'OEM'.

ODM are the people that manufacture the actual device..
OEM installs the OS, puts the harddrive in and memory and provides support.

Dell, HP, System76 (or at least the people they work through), Apple, Leveno are all 'OEM's.

ODM are places like Asus, Quanta, and Compal.

I think that Quanta is the biggest one. (for instance I beleive they make Apple's stuff)

Asus is nice because you can buy a barebones stuff directly from them, but it's a bit difficult as there is no standardization. For instance DVD drive of one type may not fit into a laptop and it's difficult to figure out what works before you buy it.

Tuxmobil has a partial matrix of relationships between OEM and ODM.
http://tuxmobil.org/laptop_oem.html

Binary-only modules not to be banned - this time

Posted Dec 15, 2006 7:48 UTC (Fri) by man_ls (guest, #15091) [Link] (1 responses)

I have found tuxmobil.org to be fairly useful, for used laptops at least.

Binary-only modules not to be banned - this time

Posted Dec 15, 2006 9:42 UTC (Fri) by Arker (guest, #14205) [Link]

Linuxlaptops.org and tuxmobil.org appear to be the same site.

R-cubed is another supplier of pre-installed Linux laptops

Posted Dec 15, 2006 17:34 UTC (Fri) by ofeeley (guest, #36105) [Link]

R-cubed sell a Core 2 Duo laptop with the Intel 945GM chipset
http://www.shoprcubed.com/proddetail.asp?prod=GW1460&...

January's Linux Journal has a very postitive review of R-cubed's (Pentium-M 760, 915G) LS1250 machine.
http://www.shoprcubed.com/proddetail.asp?prod=LS1250&...

The base machine seems to be an Asus Z33Ae.

Binary-only modules not to be banned - this time

Posted Dec 15, 2006 13:44 UTC (Fri) by DouglasJM (subscriber, #6435) [Link]

Great discussion, I'm glad that this is important to many people.

I think it may be good though to remember the point that Linus made about copy rights and telling people how they can use what is being given to them.

One of the most important things about the Linux kernel, and why it has experienced such rapid growth, is the lack of restrictions on what you can add to it and how you use it. I think that if we start telling people you can't do this or you can't do that, we run the risk of slowing the momentum that the kernel now enjoys.

What is more important, as I saw mentioned in these comments a couple times, is that we vote with our pocket books. Buy the stuff that is open, that specs are released for, and that follow the principles of open source. If we all do that to the best of our ability then we create the desire for hardware manufactures to open up their drivers.