Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Pascal Schmidt]

On Mon, 06 Oct 2003 20:50:12 +0200, you wrote in linux.kernel:

> That has no bearing on the legalities.  A version of the kernel can't
> force the GPL on a driver that works with that version of the kernel
> because you can pull that driver out and drop in another.

Okay, I can see the boundary. We still have the problem that drivers
writers have to be very careful to not copy kernel code by accident
because the kernel changes often, which creates a temptation to look
closely at in-tree drivers to see how they do things. And if a
drivers writer then produces code that is essentialy the same as is
found in the kernel, only with changed indentation and variable names,
I think we both a agree that such a driver would be a derived work.

Another problem is the fact that Linux kernel headers can contain code
in the form of macros. If a driver uses such a header, it links kernel
code with itself which can easily make it a derived work.

-- 
Ciao,
Pascal

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[David Lang]

Correct, this is why most device drivers ARE derived from the kernel and
must be GPL'd.

however it is possible to write a device driver independantly and then
interface it with the kernel without making it a derived work.

David Lang

On Mon, 6 Oct 2003, Pascal Schmidt wrote:

> Date: Mon, 6 Oct 2003 20:56:25 +0200
> From: Pascal Schmidt <der.eremit@email.de>
> To: Larry McVoy <lm@bitmover.com>
> Cc: linux-kernel@vger.kernel.org
> Subject: Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]
>
> On Mon, 06 Oct 2003 20:50:12 +0200, you wrote in linux.kernel:
>
> > That has no bearing on the legalities.  A version of the kernel can't
> > force the GPL on a driver that works with that version of the kernel
> > because you can pull that driver out and drop in another.
>
> Okay, I can see the boundary. We still have the problem that drivers
> writers have to be very careful to not copy kernel code by accident
> because the kernel changes often, which creates a temptation to look
> closely at in-tree drivers to see how they do things. And if a
> drivers writer then produces code that is essentialy the same as is
> found in the kernel, only with changed indentation and variable names,
> I think we both a agree that such a driver would be a derived work.
>
> Another problem is the fact that Linux kernel headers can contain code
> in the form of macros. If a driver uses such a header, it links kernel
> code with itself which can easily make it a derived work.
>
> --
> Ciao,
> Pascal
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>

-- 
"Debugging is twice as hard as writing the code in the first place.
Therefore, if you write the code as cleverly as possible, you are,
by definition, not smart enough to debug it." - Brian W. Kernighan

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Richard B. Johnson]

On Mon, 6 Oct 2003, Pascal Schmidt wrote:

> On Mon, 06 Oct 2003 20:50:12 +0200, you wrote in linux.kernel:
>
> > That has no bearing on the legalities.  A version of the kernel can't
> > force the GPL on a driver that works with that version of the kernel
> > because you can pull that driver out and drop in another.
>
> Okay, I can see the boundary. We still have the problem that drivers
> writers have to be very careful to not copy kernel code by accident
> because the kernel changes often, which creates a temptation to look
> closely at in-tree drivers to see how they do things. And if a
> drivers writer then produces code that is essentialy the same as is
> found in the kernel, only with changed indentation and variable names,
> I think we both a agree that such a driver would be a derived work.
>
> Another problem is the fact that Linux kernel headers can contain code
> in the form of macros. If a driver uses such a header, it links kernel
> code with itself which can easily make it a derived work.
>
> --
> Ciao,
> Pascal


Statement of the problem:

A company makes a new device that could run under Linux.
This device uses some standard gate-arrays. Because of
this, some gate-array bits need to be loaded upon startup.

The company knows that if the competition learns that a
gate-array was used, instead of an ASIC, the competition
could clone the whole device in a few weeks, thereby
stealing a few million dollars of development effort.
In fact, the thieves don't even need to buy an initial
board to copy. They just need the GPL source-code of
the driver.

The thieves need only to get a copy of the bits and
the device type of the gate-array. They then have
the entire design at-hand without ever doing any
Engineering at all. This will give the thieves an
unfair marketing advantage, allowing them to sell
the device at a cheaper amount and still make money.

For Linux to be more widely accepted by the hardware
community, there needs to be some way of protecting
the investment that the hardware companies have made.
They need to be able to hide the implementation details
from the competition. For this, they currently need to
link their GPL module-code against some secret code
that is not released to the general public.

Unfortunately, any code that can execute in the kernel
can throughly corrupt other kernel code. This has led
some to create the "Tainted" advice, etc. It may be
possible to work around the bug-reporting problem if
the module is not a disk-interface module or a screen-
interface module. Just remove the module, verify the
bug still exists, then submit a report. However, many
modules can't be removed and still have a running
system. Therefore the "Tainted" fix doesn't work.

So, instead of arguing, forever, about what's GPL
and what's not GPL, it would be real nice if the same
amount of effort was expended towards fixing the
basic problem.

The basic problem is that, in many cases, a hardware
vendor cannot divulge the inner workings of his hardware.
Even if the hardware vendor wanted to be a "nice guy"
and let everybody know how his board worked, if the
company was being publically-traded, the stock-holders
could (read would) sue and whomever released the information
could (read would) be found guilty of stealing or
divulging trade-secrets.

So, let's figure out how to protect the inner workings
of the kernel while, at the same time, protecting a
hardware vendor's engineering investment.

Cheers,
Dick Johnson
Penguin : Linux version 2.4.22 on an i686 machine (797.90 BogoMips).
            Note 96.31% of all statistics are fiction.

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Pavel Machek]

Hi!

> A company makes a new device that could run under Linux.
> This device uses some standard gate-arrays. Because of
> this, some gate-array bits need to be loaded upon startup.
> 
> The company knows that if the competition learns that a
> gate-array was used, instead of an ASIC, the competition
> could clone the whole device in a few weeks, thereby
> stealing a few million dollars of development effort.


Since when is creating compatible hw called stealing?!
If this was such a big problem, nothing prevents you
from putting ROM with those magic bits... How much is
that? _5?
-- 
				Pavel
Written on sharp zaurus, because my Velo1 broke. If you have Velo you don't need...

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Richard B. Johnson]

On Tue, 7 Oct 2003, Pavel Machek wrote:

> Hi!
>
> > A company makes a new device that could run under Linux.
> > This device uses some standard gate-arrays. Because of
> > this, some gate-array bits need to be loaded upon startup.
> >
> > The company knows that if the competition learns that a
> > gate-array was used, instead of an ASIC, the competition
> > could clone the whole device in a few weeks, thereby
> > stealing a few million dollars of development effort.
>
>
> Since when is creating compatible hw called stealing?!

When the "compatible" device is a copy.

> If this was such a big problem, nothing prevents you
> from putting ROM with those magic bits... How much is
> that? _5?

Yes it does. Market pressure. The ROM may cost US$0.50.
During the lifetime of the product, that may mean over
a million dollars in lost profit. And, if you were a
stockholder, you would not like that. Or, if you
were an employee who lost his job because the company
couldn't quite make up the cost of your salary. Every
dime saved in the production cost of a high-volume product
means several jobs saved.

> --
> 				Pavel
> Written on sharp zaurus, because my Velo1 broke. If you have Velo you don't need...
>

Cheers,
Dick Johnson
Penguin : Linux version 2.4.22 on an i686 machine (797.90 BogoMips).
            Note 96.31% of all statistics are fiction.

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[David S. Miller]

Richard, I asked the list TWICE nicely to stop this thread.

THIS MEANS YOU TOO!

If you post one more time in this thread, I am going to remove
you.

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Jamie Lokier]

Richard B. Johnson wrote:
> > Since when is creating compatible hw called stealing?!
> 
> When the "compatible" device is a copy.

The gate-array or other binary-only firmware that is distributed
alongside an open source driver is not covered by an open source license.

(We may argue about the merits of that, but it is the de facto situation).

So you, as the owner of the firmware, have _strong_ copyright
protection against that firmware being used inappropriately.
That's the legal protection.

As for technical protection, if a clone product could make "a million
dollars", the cost of extracting firmware from a closed source driver
is trivial in comparison.  It takes a skilled coder minutes or hours;
nothing compared to the difficulty of making a clone board.

Contrarily, if you and the cloner are doing it all closed source, you
are not likely to know your firmware has been copied.  OTOH if it's
the norm to produce open source drivers, you can inspect your
competition's code to see if they ripped your firmware.  (They have to
release open source too, if it's the norm, because that's what's demanded).

The real issue is that when generic chips being used, some h/w
manufacturers aren't really making hardware.  They're putting a few
widely available chips together, and offering a design.  In effect
they are becoming software companies, and we already see how the
economics of that are changing world-wide.

> > If this was such a big problem, nothing prevents you
> > from putting ROM with those magic bits... How much is
> > that? _5?
> 
> Yes it does. Market pressure. The ROM may cost US$0.50.
> During the lifetime of the product, that may mean over
> a million dollars in lost profit. And, if you were a
> stockholder, you would not like that. Or, if you
> were an employee who lost his job because the company
> couldn't quite make up the cost of your salary. Every
> dime saved in the production cost of a high-volume product
> means several jobs saved.

It is relative; the million isn't important, if you made a billion
from the product.  It's far more important to asses how the $0.50
affects your margin.  On a device costing $15 to manufacture, it
depends on whether you are selling the device to the reseller for $16
or $50.

-- Jamie

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Jamie Lokier]

Pavel Machek wrote:
> > A company makes a new device that could run under Linux.
> > This device uses some standard gate-arrays. Because of
> > this, some gate-array bits need to be loaded upon startup.
> > 
> > The company knows that if the competition learns that a
> > gate-array was used, instead of an ASIC, the competition
> > could clone the whole device in a few weeks, thereby
> > stealing a few million dollars of development effort.
> 
> Since when is creating compatible hw called stealing?!
> If this was such a big problem, nothing prevents you
> from putting ROM with those magic bits... How much is
> that? _5?

Large modern gate arrays use encrypted bitstreams, and even when not
encrypted are very hard to reverse engineer, so that's not the
problem.  It's possible, but very expensive.

Small gate arrays use non-volatile programming anyway.

The problem is blatant copying of the bitstream.

This is trivial whether it's in ROM or not, for anyone capable of
making a device, so gate-array firmware is *no excuse* for keeping the
driver code obscured, not even a _5 excuse.

-- Jamie

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[David S. Miller]

Jamie, you're on the shit list too, stop posting in this thread
or you'll be removed from the lists.

Thank you.

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Jamie Lokier]

David S. Miller wrote:
> Jamie, you're on the shit list too, stop posting in this thread
> or you'll be removed from the lists.

Sorry Dave, I didn't see your warning about this thread.

-- Jamie

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Andre Hedrick]

On Mon, 6 Oct 2003, Pascal Schmidt wrote:

> On Mon, 06 Oct 2003 20:50:12 +0200, you wrote in linux.kernel:
> 
> > That has no bearing on the legalities.  A version of the kernel can't
> > force the GPL on a driver that works with that version of the kernel
> > because you can pull that driver out and drop in another.
> 
> Okay, I can see the boundary. We still have the problem that drivers
> writers have to be very careful to not copy kernel code by accident
> because the kernel changes often, which creates a temptation to look
> closely at in-tree drivers to see how they do things. And if a
> drivers writer then produces code that is essentialy the same as is
> found in the kernel, only with changed indentation and variable names,
> I think we both a agree that such a driver would be a derived work.

You can look all you want, just can not touch.

Simiar to the red light district in Holland, it costs alot to do more
than look throught the glass.

> Another problem is the fact that Linux kernel headers can contain code
> in the form of macros. If a driver uses such a header, it links kernel
> code with itself which can easily make it a derived work.

No it can not, by only using the headers as the functional API for that
snapshot verson of the kernel release, it is the standard means for
functionality.  If the macro is require for any driver and or one in the
kernel to function, and is listed in the headers, it is generally deemed
to part of the unportected API.

Again it is very simple declare, all modules which are not GPL and reject
loading, and we can watch the death of linux as nobody will use it.  Again
who cares, because it started out as fun for a Finn in 1991, and should
never be of use or value outside of academics.

All of the code monkeys here need to be equally exploited by all.  Some of
the best marketing pitchs to date for some of the big companies who claim
to be linux friends, their slogans translated:

Use our hardware with Linux, there are a lot of suckers out there with
talnet who will gladly help you with your support issues.  This reduces
our support costs because we cans send you to join the rest of the
monkeys, Pinocchio.

Harsh and reality suck, but I could not believe this is what is being
pitched at trade shows.

Later ...

Andre

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Jamie Lokier]

Andre Hedrick wrote:
> Use our hardware with Linux, there are a lot of suckers out there with
> talnet who will gladly help you with your support issues.  This reduces
> our support costs because we cans send you to join the rest of the
> monkeys, Pinocchio.
> 
> Harsh and reality suck, but I could not believe this is what is being
> pitched at trade shows.

It is hardly surprising as plenty of Linux advocates have drawn
attention to the availability of community support.  That's one of the
purposes of Free Software.  It's considered a feature!  (Even if we
are suckers ;)

-- Jamie

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Pascal Schmidt]

On Mon, 6 Oct 2003, Andre Hedrick wrote:

> No it can not, by only using the headers as the functional API for that
> snapshot verson of the kernel release, it is the standard means for
> functionality.

Well, I don't see "standard means for functionality" mentioned anywhere
in the GPL or copyright law (though I'm no expert on that).

If a header contains a macro that expands to real code and a module
has to use that, it means that it absolutely needs that part of kernel
source code to function and then it is a derived work because it
includes GPL'ed code and would not work without it.

> If the macro is require for any driver and or one in the
> kernel to function, and is listed in the headers, it is generally deemed
> to part of the unportected API.

Says who? Who defines what is unprotected API and what is not?

> Again it is very simple declare, all modules which are not GPL and reject
> loading, and we can watch the death of linux as nobody will use it.  Again
> who cares, because it started out as fun for a Finn in 1991, and should
> never be of use or value outside of academics.

Well, silly me, I only buy hardware with open source drivers available.
I wouldn't agree that something is good and has to be done just because
it would improve Linux' "success" (I wouldn't define that to be
commercial success, either).

-- 
Ciao,
Pascal

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Andre Hedrick]

On Tue, 7 Oct 2003, Pascal Schmidt wrote:

> On Mon, 6 Oct 2003, Andre Hedrick wrote:
> 
> > No it can not, by only using the headers as the functional API for that
> > snapshot verson of the kernel release, it is the standard means for
> > functionality.
> 
> Well, I don't see "standard means for functionality" mentioned anywhere
> in the GPL or copyright law (though I'm no expert on that).

You should pay a lawyer and learn something then.

> If a header contains a macro that expands to real code and a module
> has to use that, it means that it absolutely needs that part of kernel
> source code to function and then it is a derived work because it
> includes GPL'ed code and would not work without it.

Sorry but usage of headers the functional api is an unprotectable work.
To follow up on the point of boundaries, it is abstract interface to
properly work as a driver in the given space.

Alternative view, using the MS DDK and the associated headers for running
in MicroSoft.  Because a driver uses MicroSoft Headers, does it make it a
derived work belonging to MicroSoft?

> > If the macro is require for any driver and or one in the
> > kernel to function, and is listed in the headers, it is generally deemed
> > to part of the unportected API.
> 
> Says who? Who defines what is unprotected API and what is not?

US Copyright and courts the rest of the pathetic world is lame and brain
dead and even follows our stupidity with DMCA laws also.

<see flame bait, above>

Lawyers who make up the meaning of the laws.  You should really decide to
spend some money to get an expensive education.  $10,000 USD or Euro
should be enough to rudely open your eyes to how the courts operate.
Copyright law is so strange and painful.

I know it was an expensive lesson for me to start writing checks to
lawyers, but less costly than making mistakes based on ignorance.

> > Again it is very simple declare, all modules which are not GPL and reject
> > loading, and we can watch the death of linux as nobody will use it.  Again
> > who cares, because it started out as fun for a Finn in 1991, and should
> > never be of use or value outside of academics.
> 
> Well, silly me, I only buy hardware with open source drivers available.
> I wouldn't agree that something is good and has to be done just because
> it would improve Linux' "success" (I wouldn't define that to be
> commercial success, either).

Does tongue in cheek mean anything?

Cheers,

Andre

PS one day I will figure out how to enable the spell check in pine, sigh.

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Pascal Schmidt]

Larry McVoy <lm@bitmover.com> schrieb am 07.10.03 02:57:03:

> > So you're saying the LGPL and the GPL mean the same thing for
> > libraries?  That, for instance, you can handle Qt as if it was LGPL?
> I think so, I'm afraid.  I know that this view of the law isn't what
> people think is true and the end result may well be a court case which
> tests it.

Well, for libraries, the only thing that the GPL forbids and the LGPL
allows (at least in the eyes of the FSF, grain of salt and all that) is
statically linking with the library and then distributing the resulting
program under a non-GPL license.

Fits nicely with the boundary definition you gave, because linking
statically means that the result is one program and you cannot
take it apart without wrecking it.

I think that also applies to kernel modules. Dynamically loading them
works like linking with a library dynamically  (the lib in this case being
the kernel). But statically including code into a module is like
static linking. This happens when kernel headers declare non-trivial
static inline functions or macros, and that is problematic.

All the more reason for a seperate set of cleaned up linux-abi
header, isn't it?

-- 
Ciao,
Pascal

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Pascal Schmidt]

On Mon, 06 Oct 2003 03:40:05 +0200, you wrote in linux.kernel:

> A much more obvious example than the SCM one is a device driver or a module.
> That's so cut and dried it isn't even open to debate in the eyes of the 
> law.  It's a hard and fast boundary, the GPL can't cross it no matter what
> people think or want (on either side).

Huh? How is a driver an independent work under the definition you gave?
I can't take the Linux kernel out and insert the driver into another
kernel and have it still work. Only the opposite is true - the kernel
would run without the driver, and therefore the kernel is not a derived
work of the driver and can't be subject to license terms of the driver.

The kernel doesn't have a defined interface for drivers. It changes a
lot at least during a development series. A driver is not independent from
the kernel running under it because it has to be changed quite often to
adapt to the changing internal kernel interfaces.

You can't even change the Linux kernel version you're using and have the
same driver still work for you. So by your definition, a driver is at
least in part a derived work of the kernel.

Now, if the driver has an internal abstraction layer that seperates the
kernel side of things from the real work the driver does, I would agree
that only the abstraction layer is a derived work and has to be GPL'd,
not the rest of the driver.

Most drivers don't work that way because of the additional overhead.

-- 
Ciao,
Pascal

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Larry McVoy]

On Mon, Oct 06, 2003 at 08:28:46PM +0200, Pascal Schmidt wrote:
> On Mon, 06 Oct 2003 03:40:05 +0200, you wrote in linux.kernel:
> 
> > A much more obvious example than the SCM one is a device driver or a module.
> > That's so cut and dried it isn't even open to debate in the eyes of the 
> > law.  It's a hard and fast boundary, the GPL can't cross it no matter what
> > people think or want (on either side).
> 
> Huh? How is a driver an independent work under the definition you gave?
> I can't take the Linux kernel out and insert the driver into another
> kernel and have it still work. Only the opposite is true - the kernel
> would run without the driver, and therefore the kernel is not a derived
> work of the driver and can't be subject to license terms of the driver.

The thing that is trying to cross the boundary is the kernel license
so what matters is if the thing which you believe should be GPLed is
separable or not.

> The kernel doesn't have a defined interface for drivers. It changes a
> lot at least during a development series. A driver is not independent from
> the kernel running under it because it has to be changed quite often to
> adapt to the changing internal kernel interfaces.

That has no bearing on the legalities.  A version of the kernel can't
force the GPL on a driver that works with that version of the kernel
because you can pull that driver out and drop in another.  A great example
is the eepro driver, there is Becker's version and the Intel version.
Any judge who wasn't fooled by Microsoft priced lawyers would clearly
see the boundary and make a ruling that the GPL can't cross over it.

By the way, many people here want to argue against this point of view
because they are pro GPL.  OK, fine, maybe you can change the laws
and make that stick.  I very much doubt it but let's suppose you do.
By doing that you will be supporting SCO's legal case.  If the GPL can
cross over those boundaries than so can SCO's license.  You can't have
one set of rules for you and another set of rules for them, you have to
apply them to everyone.

When you understand that you will understand more clearly why I bother to
comment on this at all.
-- 
---
Larry McVoy              lm at bitmover.com          http://www.bitmover.com/lm

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Olivier Galibert]

On Mon, Oct 06, 2003 at 11:38:57AM -0700, Larry McVoy wrote:
> That has no bearing on the legalities.  A version of the kernel can't
> force the GPL on a driver that works with that version of the kernel
> because you can pull that driver out and drop in another.  A great example
> is the eepro driver, there is Becker's version and the Intel version.
> Any judge who wasn't fooled by Microsoft priced lawyers would clearly
> see the boundary and make a ruling that the GPL can't cross over it.

So you're saying the LGPL and the GPL mean the same thing for
libraries?  That, for instance, you can handle Qt as if it was LGPL?

  OG.

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Larry McVoy]

On Mon, Oct 06, 2003 at 11:29:42PM +0200, Olivier Galibert wrote:
> On Mon, Oct 06, 2003 at 11:38:57AM -0700, Larry McVoy wrote:
> > That has no bearing on the legalities.  A version of the kernel can't
> > force the GPL on a driver that works with that version of the kernel
> > because you can pull that driver out and drop in another.  A great example
> > is the eepro driver, there is Becker's version and the Intel version.
> > Any judge who wasn't fooled by Microsoft priced lawyers would clearly
> > see the boundary and make a ruling that the GPL can't cross over it.
> 
> So you're saying the LGPL and the GPL mean the same thing for
> libraries?  That, for instance, you can handle Qt as if it was LGPL?

I think so, I'm afraid.  I know that this view of the law isn't what
people think is true and the end result may well be a court case which
tests it.

You can sort of see how the logic works.  There has to be some sort
of boundary, right?  Does anyone really think that if Linus hadn't
said that the GPL doesn't cross over to the user apps that the GPL
really would have crossed over?  So if there is a boundary concept,
how do you define what a boundary is?  Is that left to each license or
is that part of the law?  As far as I can tell, that's part of the law,
it has a concept of a boundary already.  The lawyers got a little squirmy
around this part and I got the sense that what is a boundary is not 
universally established.  But everyone seemed to think that allowing
licenses which bleed over into "unrelated" stuff is about as legal as
contracts imposing human slavery, i.e., both are not legal.

It's certainly not a done a deal, I think that sooner or later there will
be some court case that establishes the case law on which all future
cases will be based.  The people I know who care about the GPL dread
this case because the GPL has sort of had it both ways for a long time
and that can't continue.  If you want the law to be that SCO can't claim
all the IP that was built on top of the original Unix then that same law
also says that the GPL can't claim dominion over separable works either.

I can also image that the Qt people are less than thrilled with what I'm
saying because it basically invalidates the GPL-ed library or pay for a
non-GPLed version business model.

All of what I say should be taken with a grain of salt.  Yeah, I've spent
money trying to understand this and I perhaps understand more than some
people here (maybe a lot of people in this instance).  That doesn't mean
what I say is right.  If you really care, if your business depends on
it, you need to get a lawyer to work through the issues.  Right now,
I think the side with the deeper pockets would win, so even though I'm 
pointing at what I think will be the long term outcome, I'm not sure I'd
bet the farm on that.  How's that for weasel words?
-- 
---
Larry McVoy              lm at bitmover.com          http://www.bitmover.com/lm

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[David Woodhouse]

On Mon, 2003-10-06 at 11:38 -0700, Larry McVoy wrote:
> The thing that is trying to cross the boundary is the kernel license
> so what matters is if the thing which you believe should be GPLed is
> separable or not.

Forget boundaries, Larry. Consider the case of the Creosote Public
Licence to which I referred before. That one required you to bathe daily
in creosote and release _all_ your future work under the same licence;
separate works or not.

If you don't comply with the licence, you may not use the original work.
It's that simple -- whatever the requirements of the licence are, you
obey them or you don't have a licence.

In the case of the CPL, it isn't a crime for you to publish your own
non-derived work under another licence, or one day to decide not to
bathe in creosote -- but it does place you in violation of the Creosote
Public Licence and hence mean that continued use of the _original_ work
is a violation of its copyright; and therefore a criminal offence.

This is not about whether a licence _can_ demand this. We know it can --
it can demand the ritual sacrifice of your first-born, and all that
means is that if you don't agree, you don't get to use the software in
question¹.

This is about whether the GPL _does_ demand this. I believe that it
does, and that the user-space exception and the existence of the LGPL
make that fact entirely clear.

-- 
dwmw2

¹ Admittedly, incitement to murder is an offence in most countries
  but you get the point, and you still wouldn't be permitted to
  use the software if you didn't do it :)

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Andre Hedrick]

We can make up the Monica Public License too, but because it is GPL, your
added restrictions of whatever are NULL and VOID.  So Creosote Public
License, is an effective restriction that is not allowed in GPL.

All your copyright under GPL grants you copyright, all other terms are in
the license and adding anything to it is not allowed.

Adding your views in how it should be view is your opinion nothing more.
It has no legal bases and is only random noise to promote a wider grey
line and more confusion.

There is no standing legal decision in a court of law you can point to to
use as a referrence for case law or brief/opinion from the position,
officer of the court.

So lets add the Rent License, you pay rent to use the GPL code I have
published.  Anybody who has ever stored any data on an IDE/ATA/SATA device
or ATAPI attached to HOST connected to SFF cable headers of 40 pin or SFF
cable PHY headers, must pay me 0.001 cents per byte transfered regardless
of direction of success.

This on top of GPL is a usage restriction.

Have a cold shower dude, Creosote reminds me of Slick Willy.

Cheers,

Andre Hedrick
LAD Storage Consulting Group

On Tue, 7 Oct 2003, David Woodhouse wrote:

> On Mon, 2003-10-06 at 11:38 -0700, Larry McVoy wrote:
> > The thing that is trying to cross the boundary is the kernel license
> > so what matters is if the thing which you believe should be GPLed is
> > separable or not.
> 
> Forget boundaries, Larry. Consider the case of the Creosote Public
> Licence to which I referred before. That one required you to bathe daily
> in creosote and release _all_ your future work under the same licence;
> separate works or not.
> 
> If you don't comply with the licence, you may not use the original work.
> It's that simple -- whatever the requirements of the licence are, you
> obey them or you don't have a licence.
> 
> In the case of the CPL, it isn't a crime for you to publish your own
> non-derived work under another licence, or one day to decide not to
> bathe in creosote -- but it does place you in violation of the Creosote
> Public Licence and hence mean that continued use of the _original_ work
> is a violation of its copyright; and therefore a criminal offence.
> 
> This is not about whether a licence _can_ demand this. We know it can --
> it can demand the ritual sacrifice of your first-born, and all that
> means is that if you don't agree, you don't get to use the software in
> question¹.
> 
> This is about whether the GPL _does_ demand this. I believe that it
> does, and that the user-space exception and the existence of the LGPL
> make that fact entirely clear.
> 
> -- 
> dwmw2
> 
> ¹ Admittedly, incitement to murder is an offence in most countries
>   but you get the point, and you still wouldn't be permitted to
>   use the software if you didn't do it :)
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Roman Zippel]

Hi,

On Tue, 7 Oct 2003, Andre Hedrick wrote:

> We can make up the Monica Public License too, but because it is GPL, your
> added restrictions of whatever are NULL and VOID.

Could you please explain about what "added restrictions" you're talking 
about? Let's actually look at the GPL, which states "You may not impose 
any further restrictions on the recipients' exercise of the rights granted 
herein.", a bit earlier we find "Activities other than copying, 
distribution and modification are not covered by this License".
So how exactly does EXPORT_SYMBOL_GPL restrict you in these activities?

bye, Roman

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Andre Hedrick]

Roman,

You got my point exactly, keep the issue of license to GPL and not muddy
the waters with a license other than GPL as an example.

David's "Creosote Public License" is an example how an author using such a
license could require the enduser to conform to his/her wishes,
regardless.

His "Creosote Public License" adds restrictions not permitted or covered
by the scope of GPL.  So generating examples of how that license empowers
the author to impose extra restrictions and then saying it is equal to GPL
by association is not helping to clear the mudd.

Sorry if I was not clear about the direction of the message.

Cheers,

Andre Hedrick
LAD Storage Consulting Group

On Tue, 7 Oct 2003, Roman Zippel wrote:

> Hi,
> 
> On Tue, 7 Oct 2003, Andre Hedrick wrote:
> 
> > We can make up the Monica Public License too, but because it is GPL, your
> > added restrictions of whatever are NULL and VOID.
> 
> Could you please explain about what "added restrictions" you're talking 
> about? Let's actually look at the GPL, which states "You may not impose 
> any further restrictions on the recipients' exercise of the rights granted 
> herein.", a bit earlier we find "Activities other than copying, 
> distribution and modification are not covered by this License".
> So how exactly does EXPORT_SYMBOL_GPL restrict you in these activities?
> 
> bye, Roman
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Roman Zippel]

Hi,

On Tue, 7 Oct 2003, Andre Hedrick wrote:

> You got my point exactly, keep the issue of license to GPL and not muddy
> the waters with a license other than GPL as an example.

Fine, but you didn't answer my questions at all and you're the one 
complaining about EXPORT_SYMBOL_GPL in first place, so here are the 
questions again, in case you deleted the original mail already:

Could you please explain about what "added restrictions" you're talking 
about? Let's actually look at the GPL, which states "You may not impose 
any further restrictions on the recipients' exercise of the rights granted 
herein.", a bit earlier we find "Activities other than copying, 
distribution and modification are not covered by this License".
So how exactly does EXPORT_SYMBOL_GPL restrict you in these activities?

bye, Roman

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Andre Hedrick]

"EXPORT_SYMBOL_GPL" is a license enforcement tool, period.

Any driver using a caller wrappered by "EXPORT_SYMBOL_GPL" will have the
function fail quietly and restrict functionality.  Restricting
functionality is a means to disable drivers period.  These very same
functions in question are required for any other driver to operate.

That makes them part of the API for operational compatibility.
The API is described exclusively by the snapshot version of a given
kernel.

So under copyright case law, where GPL has never been, there are rulings
and history describing what an API is, how it can not be protected, and
determinations of fair use based on legal rulings.

Now if one is forced to disable EXPORT_SYMBOL_GPL and redistribute the
modified kernel source as a single object, there is nothing anybody can do
to stop that process.

A simple solution is to publish an EXPORT_SYMBOL shim so one would not
be subject to the issues of distribution clauses.  The SHIM is
GPL-compatable and it provides clear exports of the API snapshot.

Separtion of the logical .c shim module driver and the decoupled .h is
making the boundary clear.

However this is a was of time because neither you (best guess) and the
majority here have ever paid to have the license explained and how it is
subject to actual copyright laws.  In the past I was one of the folks
shouting opinions not facts.  The facts are rude eye openers.

Well David Miller is going to stomp the thread so lets do it for him?

Cheers,


Andre Hedrick
LAD Storage Consulting Group

On Tue, 7 Oct 2003, Roman Zippel wrote:

> Hi,
> 
> On Tue, 7 Oct 2003, Andre Hedrick wrote:
> 
> > You got my point exactly, keep the issue of license to GPL and not muddy
> > the waters with a license other than GPL as an example.
> 
> Fine, but you didn't answer my questions at all and you're the one 
> complaining about EXPORT_SYMBOL_GPL in first place, so here are the 
> questions again, in case you deleted the original mail already:
> 
> Could you please explain about what "added restrictions" you're talking 
> about? Let's actually look at the GPL, which states "You may not impose 
> any further restrictions on the recipients' exercise of the rights granted 
> herein.", a bit earlier we find "Activities other than copying, 
> distribution and modification are not covered by this License".
> So how exactly does EXPORT_SYMBOL_GPL restrict you in these activities?
> 
> bye, Roman
> 

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Roman Zippel]

Hi,

On Tue, 7 Oct 2003, Andre Hedrick wrote:

> "EXPORT_SYMBOL_GPL" is a license enforcement tool, period.
> 
> Any driver using a caller wrappered by "EXPORT_SYMBOL_GPL" will have the
> function fail quietly and restrict functionality.

If I understand you correctly, you cannot tell me how EXPORT_SYMBOL_GPL 
is an additional restriction according to the GPL.
OTOH what's wrong with enforcing a license?

>  Restricting
> functionality is a means to disable drivers period.  These very same
> functions in question are required for any other driver to operate.
> 
> That makes them part of the API for operational compatibility.
> The API is described exclusively by the snapshot version of a given
> kernel.

What gave you the right to use the API or the implementation of this API 
in first place?

> So under copyright case law, where GPL has never been,

Could you please explain, what you're trying to say here?

> there are rulings
> and history describing what an API is, how it can not be protected, and
> determinations of fair use based on legal rulings.
> 
> Now if one is forced to disable EXPORT_SYMBOL_GPL and redistribute the
> modified kernel source as a single object, there is nothing anybody can do
> to stop that process.

Why would you be forced to disable EXPORT_SYMBOL_GPL? You can do that of 
course as long as you respect the conditions in the GPL, but that wouldn't 
require modifying EXPORT_SYMBOL_GPL.

> A simple solution is to publish an EXPORT_SYMBOL shim so one would not
> be subject to the issues of distribution clauses.  The SHIM is
> GPL-compatable and it provides clear exports of the API snapshot.
> 
> Separtion of the logical .c shim module driver and the decoupled .h is
> making the boundary clear.

Could you please explain how your binary module can be "reasonably 
considered independent", so that you don't have to distribute it under the 
GPL? How are boundaries of any importance, if you not allowed to use the 
GPL part in first place? Alternatively what gives you the right to ignore 
the will of the copyright owner and distribute the GPL part under your own 
conditions?

> However this is a was of time because neither you (best guess) and the
> majority here have ever paid to have the license explained and how it is
> subject to actual copyright laws.  In the past I was one of the folks
> shouting opinions not facts.  The facts are rude eye openers.

Well, what I have seen so far is that your logic stinks mightily and you 
are twisting reality to your own needs and on this way showing absolutely 
no respect for the work of others.

bye, Roman

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[David S. Miller]

Please guys, stop this thread as I kindly asked earlier
this evening.

I hate being truant officer, I in fact hate it.  So please
just be reasonable.

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[David S. Miller]

This discussion is more and more about licenses plus law and less and
less about the kernel in particular.

Please folks, take it somewhere else.

Thanks.

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Larry McVoy]

On Tue, Oct 07, 2003 at 09:40:37AM +0100, David Woodhouse wrote:
> This is not about whether a licence _can_ demand this. We know it can --
> it can demand the ritual sacrifice of your first-born, and all that
> means is that if you don't agree, you don't get to use the software in
> question??.

No, that's the point you have been missing all along.  Any unenforceable 
clause is null and void.  Read section 7 of the GPL and a little law.
You can't enforce unenforceable clauses by revoking the license.
-- 
---
Larry McVoy              lm at bitmover.com          http://www.bitmover.com/lm

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 776 bytes --]

On Tue, 07 Oct 2003 07:16:36 PDT, Larry McVoy said:

> No, that's the point you have been missing all along.  Any unenforceable 
> clause is null and void.  Read section 7 of the GPL and a little law.
> You can't enforce unenforceable clauses by revoking the license.

Actually, most fairly complicated contracts (more than just 2-3 clauses) have a
separability clause because if you don't, then the whole contract DOES go poof
if you have an unenforceable clause - so without one, the license goes away,
revoking itself.

And sure enough, section 7 has:

If any portion of this section is held invalid or unenforceable under
any particular circumstance, the balance of the section is intended to
apply and the section as a whole is intended to apply in other
circumstances



[-- Attachment #2: Type: application/pgp-signature, Size: 226 bytes --]

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[David Woodhouse]

On Mon, 2003-10-06 at 20:28 +0200, Pascal Schmidt wrote:
> Now, if the driver has an internal abstraction layer that seperates the
> kernel side of things from the real work the driver does, I would agree
> that only the abstraction layer is a derived work and has to be GPL'd,
> not the rest of the driver.

> Most drivers don't work that way because of the additional 
> overhead.

And because making such distinction is pointless, since the GPL'd
wrapper and the core driver would not be distributed 'as separate works'
but rather 'as part of a whole which is a work based on the Program',
where the Program in this case is the GPL'd wrapper part. 

Hence under the terms of the final paragraph of section 2 of the GPL,
the code of the driver would also have to be released under the same
terms.

-- 
dwmw2

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[walt]

Andre Hedrick wrote:
> You are so young and fresh to the game, it is cute.
> 
> http://www.gcom.com/home/support/whitepapers/linux-gnu-license.html
> 
> Try doing a little more checking of your history Rob.
> 
> The position was set in 1995 and to many who objected let it unchallanged.
> Now regardless if you knew of this information or not when you entered
> into kernel development, tough!  I did not know about it and have to suck
> it up that binary modules are permitted, period.  So if other people can
> make money, then I will too, and encourage others to persue also...

Hi Andre,

Since I was bold enough to criticize your writing not long ago, I want
to say now that I had absolutely no trouble understanding this post.  It
was a pleasure to read, and I thank you for taking the time and effort
to make it that way.  The improvement is dramatic, and it really helps.

Keep up the good work!

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Pascal Schmidt]

On Sun, 14 Sep 2003 09:20:08 +0200, you wrote in linux.kernel:

> If one reads ./include/linux/module.h
> It clearly states any license is acceptable.

The problem is that the kernel has many copyright holders for different
parts of the code, so what the author of module.h wrote can well be
his own opinion but not that of others. Also, if modules are derived
works under the terms of the GPL, the header file cannot change that
fact, no matter what is says exactly.

Once again, it's all up to the lawyers to decide.

-- 
Ciao,
Pascal

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Stan Bubrouski]

*END THIS THREAD*

Guys,

This arguing is not accomplishing anything.  Insults
accomplish nothing.  Grow up.  Take it to private
mail, this rubbbish is unnecessary.

-sb

Re: People, not GPL [was: Re: Driver Model]

[Erik Andersen]

On Sat Sep 13, 2003 at 08:36:36PM -0700, Andre Hedrick wrote:
> Wow ... Does "Original Work" have meaning?
> 
> Does an "Original Work" using only the standard kernel API headers to
> interface mean it is a derived work?  You better go find a new lawyer.

You seem to be somewhat confused as to who needs a lawyer.  I'm
not the one asking this question.   I am also not the one trying
to make a closed source binary only product that runs within the
context of the Linux kernel, and then complaining that the GPL
wackos are ruining my business...  It seems to be that doing such
a thing would be a really stupid business model.

As I recall it is the One True(tm) iSCSI stack you are working
on, right?

> "fair usage" of .h files as the API is standard.
> 
> Using any .c or kernel C code is a NO NO.

I invite you to read the COPYING file included in each and every
kernel tarball.  There is exactly ONE exception granted in the
linux kernel copyright:

    This copyright does *not* cover user programs that use kernel
    services by normal system calls - this is merely considered
    normal use of the kernel, and does *not* fall under the
    heading of "derived work".

All the noise in the world about other exceptions is precisely
that, since the license granting use of the Linux kernel does
not contain any additional provisions.

Anything that can be identified as a "user program" that "use[s]
kernel services by normal system calls" is, by virtue of the above
license grant, doing so with permission and is therefore within
its rights.  So you can make all the closed source user space
only One True(tm) iSCSI stacks you want.

Anything that is not a "user program" (and I think everyone can
agree a kernel module is not a "user program") is therefore a
derivitive work.

Anything that is linked into the kernel (and I think everyone can
agree a kernel module is linked into the kernel) and is therefore
interfacing with kernel internals, rather than using "kernel
services by normal system calls" is therefore a derivitive work.

Laugh at people, mock people, rant, rave, wantever you want.
When you are done making noise, please have your laywer explain
how a closed source binary only product that runs within the
context of the Linux kernel is not a derivitive work, per the
very definition given in the kernel COPYING file that grants you
your limited rights for copying, distribution and modification,

 -Erik

--
Erik B. Andersen             http://codepoet-consulting.com/
--This message was written using 73% post-consumer electrons--

freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Andre Hedrick]

Pretty Boy,

It is coming and the intent is to return all the stolen symbols.
It is free for anyone to use and enjoy the usage of Linux once again.
So everyone get in line and SUE ME for GPL'ed drivers.


/*
 * Original copyright notice:
 *
 * linux/kernel/freed_symbols.c
 *
 * Copyright (C) 2001-2003              Linux ATA Development
 *                                      Andre Hedrick <andre@linux-ide.org>
 *
 * GPL v2 and only version 2.
 */

/*
 * kernel/signal.c:EXPORT_SYMBOL_GPL(dequeue_signal);
 * returned to kernel API
 */
int freed_dequeue_signal(sigset_t *mask, siginfo_t *info)
{
        return dequeue_signal(mask, info);
}

EXPORT_SYMBOL(freed_dequeue_signal);

/*
 * was kernel/context.c
 *
 * kernel/workqueue.c:EXPORT_SYMBOL_GPL(create_workqueue);
 * kernel/workqueue.c:EXPORT_SYMBOL_GPL(queue_work);
 * kernel/workqueue.c:EXPORT_SYMBOL_GPL(queue_delayed_work);
 * kernel/workqueue.c:EXPORT_SYMBOL_GPL(flush_workqueue);
 * kernel/workqueue.c:EXPORT_SYMBOL_GPL(destroy_workqueue);
 * returned to kernel API
 */

struct workqueue_struct *freed_create_workqueue(const char *name)
{
        return create_workqueue(name);
}

int freed_queue_work(struct workqueue_struct *wq, struct work_struct *work)
{
        return queue_work(wq, work);
}

int freed_queue_delayed_work(struct workqueue_struct *wq,
                struct work_struct *work, unsigned long delay)
{
        return queue_delayed_work(wq, work, delay);
}

void freed_flush_workqueue(struct workqueue_struct *wq)
{
        flush_workqueue(wq);
}

void freed_destroy_workqueue(struct workqueue_struct *wq)
{
        destroy_workqueue(wq);
}

EXPORT_SYMBOL(freed_create_workqueue);
EXPORT_SYMBOL(freed_queue_work);
EXPORT_SYMBOL(freed_queue_delayed_work);
EXPORT_SYMBOL(freed_flush_workqueue);
EXPORT_SYMBOL(freed_destroy_workqueue);


static int freed_symbols_ioctl (struct inode *inode, struct file *file, unsigned int cmd, unsigned long arg)
{
        if (!capable(CAP_SYS_ADMIN))
                return -EACCES;

        switch (cmd) {
                default:
                        return -EINVAL;
        }
        return 0;
}

static int freed_symbols_open (struct inode *inode, struct file *file)
{
        return 0;
}

static int freed_symbols_release (struct inode *inode, struct file *file)
{
        return 0;
}

static struct file_operations freed_symbols_fops = {
        owner:          THIS_MODULE,
        ioctl:          freed_symbols_ioctl,
        open:           freed_symbols_open,
        release:        freed_symbols_release,
};

static struct miscdevice freed_symbols_dev = {  FREED_SYMBOLS_MINOR,
                                                "freed_symbols",
                                                &freed_symbols_fops };

static void __exit freed_symbols_exit (void)
{
        printk(KERN_INFO "freed_symbols_module: unloaded.\n");
//      remove_proc_entry("driver/freed_symbols", NULL);
        misc_deregister(&freed_symbols);
}

int __init freed_symbols_init (void)
{
        printk(KERN_INFO "freed_symbols_module: loaded.\n");
        misc_register(&freed_symbols_dev);
//      create_proc_read_entry("driver/freed_symbols", 0, 0, freed_symbols_read_proc, NULL);
        return 0;
}

MODULE_LICENSE("GPL");
module_init(freed_symbols_init);
module_exit(freed_symbols_exit);


-------------------------------------------

Cheers,

Andre Hedrick
LAD Storage Consulting Group

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Erik Andersen]

On Sat Sep 13, 2003 at 09:58:41PM -0700, Andre Hedrick wrote:
> 
> Pretty Boy,
> 
> It is coming and the intent is to return all the stolen symbols.
> It is free for anyone to use and enjoy the usage of Linux once again.
> So everyone get in line and SUE ME for GPL'ed drivers.

Do whatever you want.  Its your life.  Laugh at people, mock
people, rant, rave, violtate licenses, wantever you want.

When you are done making noise, please explain how a closed
source binary only product that runs within the context of the
Linux kernel is not a derivitive work, per the very definition
given in the kernel COPYING file that grants you your limited
rights for copying, distribution and modification,

 -Erik

--
Erik B. Andersen             http://codepoet-consulting.com/
--This message was written using 73% post-consumer electrons--

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Andre Hedrick]

Erik,

Explain how a symbol in 2.4 which was EXPORT_SYMBOL is now
EXPORT_SYMBOL_GPL in 2.6 ?

When you can explain why the API for functionallity in 2.4 is ripped off
like an old lady's purse by a two-bit punk and made nojn-functional in 2.6
you may have a point.

But, you know what, I don't give a damn (DGD).

It is wrong and the original intent when it was discussed was for "NEW
SYMBOLS ONLY".  But if distros can add in Symbols for code that does not
exist in the tree, why can't people change them?

But you have a nice day, and do you need a mail address for that letter
you want to send me?  Please make it on heavy stock, you need some fiber
in your diet.

Cheers,

Andre Hedrick
LAD Storage Consulting Group

On Sat, 13 Sep 2003, Erik Andersen wrote:

> On Sat Sep 13, 2003 at 09:58:41PM -0700, Andre Hedrick wrote:
> > 
> > Pretty Boy,
> > 
> > It is coming and the intent is to return all the stolen symbols.
> > It is free for anyone to use and enjoy the usage of Linux once again.
> > So everyone get in line and SUE ME for GPL'ed drivers.
> 
> Do whatever you want.  Its your life.  Laugh at people, mock
> people, rant, rave, violtate licenses, wantever you want.
> 
> When you are done making noise, please explain how a closed
> source binary only product that runs within the context of the
> Linux kernel is not a derivitive work, per the very definition
> given in the kernel COPYING file that grants you your limited
> rights for copying, distribution and modification,
> 
>  -Erik
> 
> --
> Erik B. Andersen             http://codepoet-consulting.com/
> --This message was written using 73% post-consumer electrons--
> 

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Erik Andersen]

On Sat Sep 13, 2003 at 10:32:38PM -0700, Andre Hedrick wrote:
> 
> Erik,
> 
> Explain how a symbol in 2.4 which was EXPORT_SYMBOL is now
> EXPORT_SYMBOL_GPL in 2.6 ?
> 
> When you can explain why the API for functionallity in 2.4 is ripped off
> like an old lady's purse by a two-bit punk and made nojn-functional in 2.6
> you may have a point.

It doesn't matter what the symbol is called.  I personally agree
with you on this one point -- changing the symbols to use
EXPORT_SYMBOL_GPL type naming is deeply stupid.  

I think it is stupid because by implication, it suggests that any
exported symbols lacking such tags are somehow NOT under the GPL.

Per the COPYING file included with each and every copy of the
kernel, Linux is licensed under the GPL.  There are no provisions
in the linux kernel COPYING statement allowing non-GPL compatible
binary only closed source kernel modules.

You are therefore, entitled to abide by the precise terms and
conditions for copying, distribution and modification for the
Linux kernel.  This entitles you to change symbol names to
whatever makes you feel happy.

But you are also _required_ to abide by the precise terms and
conditions for copying, distribution and modification for the
Linux kernel, which stipulates that unless your code is a "user
[program] that [uses] kernel services by normal system calls", it
is a derived work and therefore must abide by the terms of the
GPL.

Creating and loading such a symbol renaming module is certainly
something you are entitled to do.  Using that module for
circumvention of an "effective technological measure" that
"effectively protects a right of a copyright owner ... in the
ordinary course of its operation...." could certainly open you to
legal action here in the USA.  I do not hold copyright on any of
the symbols in question, but someone does, and if they do not take
kindly to your circumvention device....

But the DMCA issues are merely an aside to the fundamental
problem.  A problem you have avoided in this thread with
gratuitous ad hominem attacks, with the "but Billy did it
first" defence, and similar nonsence.

When you are done making noise, please explain how a closed
source binary only product that runs within the context of the
Linux kernel is not a derivitive work, per the very definition
given in the kernel COPYING file that grants you your limited
rights for copying, distribution and modification.

 -Erik

--
Erik B. Andersen             http://codepoet-consulting.com/
--This message was written using 73% post-consumer electrons--

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Andre Hedrick]

Erik,

When you can answer why binary only modules are allowed to load regardless
of symbol usage, I will answer your question.  Since binary only modules
are allowed to load, this means they are intended to function.  If they
are intended to function, they must use the API for standard operations.

Removing the abilty for standard functionality means one does not want
them to function?  So if the kernel GPL-ONLY wonder blunders are thinking,
why let them load at all?

Clearly there is a "tainting" process?

If one can detect taint, one can reject loading.

If one reads ./include/linux/module.h

It clearly states any license is acceptable.

So when you can explain about "yes you can but not here", I will explain
the simple rules of how copyright and header files are to be used.

Now what does this have to do with a "freed_symbols" project?

Simple, it restores the symbols back to their original state before
everybody had a whining feast.  When they are restored back to original
state then simple usage of the headers compliance to the "unprotectable
interface" is normalized.  Go look up 1991 copyright rulings wrt SEGA.

It is all or nothing but both ways is not allowed, what do you think, we
all live in San Francisco?

Cheers,

Andre Hedrick
LAD Storage Consulting Group

On Sun, 14 Sep 2003, Erik Andersen wrote:

> On Sat Sep 13, 2003 at 10:32:38PM -0700, Andre Hedrick wrote:
> > 
> > Erik,
> > 
> > Explain how a symbol in 2.4 which was EXPORT_SYMBOL is now
> > EXPORT_SYMBOL_GPL in 2.6 ?
> > 
> > When you can explain why the API for functionallity in 2.4 is ripped off
> > like an old lady's purse by a two-bit punk and made nojn-functional in 2.6
> > you may have a point.
> 
> It doesn't matter what the symbol is called.  I personally agree
> with you on this one point -- changing the symbols to use
> EXPORT_SYMBOL_GPL type naming is deeply stupid.  
> 
> I think it is stupid because by implication, it suggests that any
> exported symbols lacking such tags are somehow NOT under the GPL.
> 
> Per the COPYING file included with each and every copy of the
> kernel, Linux is licensed under the GPL.  There are no provisions
> in the linux kernel COPYING statement allowing non-GPL compatible
> binary only closed source kernel modules.
> 
> You are therefore, entitled to abide by the precise terms and
> conditions for copying, distribution and modification for the
> Linux kernel.  This entitles you to change symbol names to
> whatever makes you feel happy.
> 
> But you are also _required_ to abide by the precise terms and
> conditions for copying, distribution and modification for the
> Linux kernel, which stipulates that unless your code is a "user
> [program] that [uses] kernel services by normal system calls", it
> is a derived work and therefore must abide by the terms of the
> GPL.
> 
> Creating and loading such a symbol renaming module is certainly
> something you are entitled to do.  Using that module for
> circumvention of an "effective technological measure" that
> "effectively protects a right of a copyright owner ... in the
> ordinary course of its operation...." could certainly open you to
> legal action here in the USA.  I do not hold copyright on any of
> the symbols in question, but someone does, and if they do not take
> kindly to your circumvention device....
> 
> But the DMCA issues are merely an aside to the fundamental
> problem.  A problem you have avoided in this thread with
> gratuitous ad hominem attacks, with the "but Billy did it
> first" defence, and similar nonsence.
> 
> When you are done making noise, please explain how a closed
> source binary only product that runs within the context of the
> Linux kernel is not a derivitive work, per the very definition
> given in the kernel COPYING file that grants you your limited
> rights for copying, distribution and modification.
> 
>  -Erik
> 
> --
> Erik B. Andersen             http://codepoet-consulting.com/
> --This message was written using 73% post-consumer electrons--
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Andre Hedrick]

On Sun, 14 Sep 2003, Erik Andersen wrote:

> On Sat Sep 13, 2003 at 10:32:38PM -0700, Andre Hedrick wrote:
> > 
> > Erik,
> > 
> > Explain how a symbol in 2.4 which was EXPORT_SYMBOL is now
> > EXPORT_SYMBOL_GPL in 2.6 ?
> > 
> > When you can explain why the API for functionallity in 2.4 is ripped off
> > like an old lady's purse by a two-bit punk and made nojn-functional in 2.6
> > you may have a point.
> 
> It doesn't matter what the symbol is called.  I personally agree
> with you on this one point -- changing the symbols to use
> EXPORT_SYMBOL_GPL type naming is deeply stupid.  
> 
> I think it is stupid because by implication, it suggests that any
> exported symbols lacking such tags are somehow NOT under the GPL.
> 
> Per the COPYING file included with each and every copy of the
> kernel, Linux is licensed under the GPL.  There are no provisions
> in the linux kernel COPYING statement allowing non-GPL compatible
> binary only closed source kernel modules.

See above, remove the EXPORT_SYMBOL_GPL crap and then there is no need to
alter or change anything.  This restores the original API.


> You are therefore, entitled to abide by the precise terms and
> conditions for copying, distribution and modification for the
> Linux kernel.  This entitles you to change symbol names to
> whatever makes you feel happy.
> 
> But you are also _required_ to abide by the precise terms and
> conditions for copying, distribution and modification for the
> Linux kernel, which stipulates that unless your code is a "user
> [program] that [uses] kernel services by normal system calls", it
> is a derived work and therefore must abide by the terms of the
> GPL.

Nice, but "derived work" is bogus and we all know it.

> Creating and loading such a symbol renaming module is certainly
> something you are entitled to do.  Using that module for
> circumvention of an "effective technological measure" that
> "effectively protects a right of a copyright owner ... in the
> ordinary course of its operation...." could certainly open you to
> legal action here in the USA.  I do not hold copyright on any of
> the symbols in question, but someone does, and if they do not take
> kindly to your circumvention device....

See above, which you agree is stupid, make EXPORT_SYMBOL_GPL go away and
nobody has to deal with API issues.

> But the DMCA issues are merely an aside to the fundamental
> problem.  A problem you have avoided in this thread with
> gratuitous ad hominem attacks, with the "but Billy did it
> first" defence, and similar nonsence.

Nope, you made threats.

> When you are done making noise, please explain how a closed
> source binary only product that runs within the context of the
> Linux kernel is not a derivitive work, per the very definition
> given in the kernel COPYING file that grants you your limited
> rights for copying, distribution and modification.

See above again, nobody has to do anything if the API is restored to it
original format.  Thus no changes, no modifications.  All of your points
are void.


Cheers,

Andre

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Erik Andersen]

On Sun Sep 14, 2003 at 12:10:27AM -0700, Andre Hedrick wrote:
> > When you are done making noise, please explain how a closed
> > source binary only product that runs within the context of the
> > Linux kernel is not a derivitive work, per the very definition
> > given in the kernel COPYING file that grants you your limited
> > rights for copying, distribution and modification.
> 
> See above again, nobody has to do anything if the API is restored to it
> original format.  Thus no changes, no modifications.  All of your points
> are void.

Truly a dizzying intellect!  All my points are void and it is ok
to load binary only modules in the Linux kernel without releasing
source for the derivitive work.  And the reason why it is ok to
thus violate the Linux kernel licence is because....  

Oh, I guess you forgot the part where you explain why this is
legal.  Sorry, but The Great Andre has Spoken isn't good enough. 
Sorry, I'm not going to ignore the man behind the curtain.

When you are done making noise, please explain how a closed
source binary only product that runs within the context of the
Linux kernel is not a derivitive work and therefore not subject
to the terms of the GPL, per the definition given in the kernel
COPYING file that grants you your limited rights for copying,
distribution and modification.

 -Erik

--
Erik B. Andersen             http://codepoet-consulting.com/
--This message was written using 73% post-consumer electrons--

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Henning P. Schmiedehausen]

Erik Andersen <andersen@codepoet.org> writes:

>When you are done making noise, please explain how a closed
>source binary only product that runs within the context of the
>Linux kernel is not a derivitive work and therefore not subject
>to the terms of the GPL, per the definition given in the kernel
>COPYING file that grants you your limited rights for copying,
>distribution and modification.

"Because Linus said so".

	Regards
		Henning

-- 
Dipl.-Inf. (Univ.) Henning P. Schmiedehausen          INTERMETA GmbH
hps@intermeta.de        +49 9131 50 654 0   http://www.intermeta.de/

Java, perl, Solaris, Linux, xSP Consulting, Web Services 
freelance consultant -- Jakarta Turbine Development  -- hero for hire

"Dominate!! Dominate!! Eat your young and aggregate! I have grotty silicon!" 
      -- AOL CD when played backwards  (User Friendly - 200-10-15)

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Erik Andersen]

On Mon Sep 15, 2003 at 12:17:37AM +0000, Henning P. Schmiedehausen wrote:
> Erik Andersen <andersen@codepoet.org> writes:
> 
> >When you are done making noise, please explain how a closed
> >source binary only product that runs within the context of the
> >Linux kernel is not a derivitive work and therefore not subject
> >to the terms of the GPL, per the definition given in the kernel
> >COPYING file that grants you your limited rights for copying,
> >distribution and modification.
> 
> "Because Linus said so".

It does not say "Because Linus said so" in the Linux kernel
COPYING file, which is the only official document that grants
legal permission to copy, distribute and/or modify the kernel.

And even if Linus says so, what about Alan Cox, David S. Miller,
Al Viro, Andrea Arcangeli, Jens Axboe, Donald Becker, Andries
Brouwer, Jeff Garzik, Dave Jones, Russell King, Rik van Riel,
Rusty Russell, Ted Ts'o, Stephen Tweedie, etc, etc, etc?  What do
they say?  After all, the Linux kernel ceased to be a
one-man-show well over 10 years ago.  I know I have personally
submitted all my patches to the Linux kernel per the GPL, not
some imagined "GPL + kernel module exceptions" license.

If Linus wanted to say "I'm relicensing Linux under the Microsoft
EULA effective immediately", he is certainly entitled to
relicence the bits he personally wrote, but nothing more.
Similarly, if Linus wants to say the kernel allows binary only
kernel modules, he can certainly say as much -- free speech
entitles him to say whatever he wants.  But he has no authority
to relicense the bits of code I wrote, or the code anyone else
wrote, without their express permission.  I have never been asked
to agree to some other kernel license, and to my knowledge,
neither has anyone else.  Therefore the license as stated in the
kernel COPYING file is in effect.

To change the license for the entire kernel would require asking
every kernel contributor of substance to agree to new licensing
terms.  I'm sure some people (such as Andre) would be overjoyed
to have a "GPL + module exceptions" license made official.  And
as you might imagine, others would be rather less enthusiastic.
And still others, such as Leonard Zubkoff may find it difficult
to make posthumous licensing decisions.

But until such an official re-licensing effort is undertaken and
sucessfully completed, the Linux kernel is and will remain
licensed under the bog standard GPL with the one exception that
"user programs that use kernel services by normal system calls"
and not derivitive works -- i.e. the licensing terms specified
in the linux kernel COPYING file.

 -Erik

--
Erik B. Andersen             http://codepoet-consulting.com/
--This message was written using 73% post-consumer electrons--

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Nick Piggin]

Erik Andersen wrote:

>On Mon Sep 15, 2003 at 12:17:37AM +0000, Henning P. Schmiedehausen wrote:
>
>>Erik Andersen <andersen@codepoet.org> writes:
>>
>>
>>>When you are done making noise, please explain how a closed
>>>source binary only product that runs within the context of the
>>>Linux kernel is not a derivitive work and therefore not subject
>>>to the terms of the GPL, per the definition given in the kernel
>>>COPYING file that grants you your limited rights for copying,
>>>distribution and modification.
>>>
>>"Because Linus said so".
>>
>
>It does not say "Because Linus said so" in the Linux kernel
>COPYING file, which is the only official document that grants
>legal permission to copy, distribute and/or modify the kernel.
>

How about taking the "GPL exported" symbols just as notes which show
their usage is not considered a derivative work by the copyright 
holders? This is what I always thought it was there for. This of course
means someone can't simply add that "note" and expect it to change the
way the copyright holders think of their work.

I don't expect this would do much in court, but isn't that for binary
driver people to worry about because it is granting more rights than
the GPL allows. Maybe it could be written in legalease to give them
more confidence. And it lets open source developers know where they
stand.

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Rob Landley]

On Monday 15 September 2003 00:57, Erik Andersen wrote:
> On Mon Sep 15, 2003 at 12:17:37AM +0000, Henning P. Schmiedehausen wrote:
> > Erik Andersen <andersen@codepoet.org> writes:
> > >When you are done making noise, please explain how a closed
> > >source binary only product that runs within the context of the
> > >Linux kernel is not a derivitive work and therefore not subject
> > >to the terms of the GPL, per the definition given in the kernel
> > >COPYING file that grants you your limited rights for copying,
> > >distribution and modification.
> >
> > "Because Linus said so".
>
> It does not say "Because Linus said so" in the Linux kernel
> COPYING file, which is the only official document that grants
> legal permission to copy, distribute and/or modify the kernel.

Linus clearly and publicly stated his position on binary only kernel modules 
almost exactly one year ago:

http://groups.google.com/groups?selm=Pine.LNX.4.44.0210170958340.6739-100000%40home.transmeta.com.lucky.linux.kernel

He basically said there IS no module exception to the GPL, it's just a 
question of what is and is not a derived work.

The kernel developers have marked up portions of the API to indicate "we 
consider anything that needs to access this deeply internal bit to be a 
derived work, hence subject to the GPL".  That's what GPL_ONLY _means_.  
Needing to re-export that therefore opens you up to a lawsuit.  (Whether you 
can defend yourself in court from that lawsuit is always an open question, 
but by adding GPL_ONLY markup the developers made their intent much more 
clear, which is unlikely to help you convince a judge of your interpretation 
if you explicitly undo that markup and then claim the license doesn't apply 
to you...)

Here's the relevant section of the above posting from Linus:

-----

I will re-iterate my stance on the GPL and kernel modules:

  There is NOTHING in the kernel license that allows modules to be 
  non-GPL'd. 

  The _only_ thing that allows for non-GPL modules is copyright law, and 
  in particular the "derived work" issue. A vendor who distributes non-GPL 
  modules is _not_ protected by the module interface per se, and should 
  feel very confident that they can show in a court of law that the code 
  is not derived.

  The module interface has NEVER been documented or meant to be a GPL 
  barrier. The COPYING clearly states that the system call layer is such a 
  barrier, so if you do your work in user land you're not in any way 
  beholden to the GPL. The module interfaces are not system calls: there 
  are system calls used to _install_ them, but the actual interfaces are
  not.

  The original binary-only modules were for things that were pre-existing 
  works of code, ie drivers and filesystems ported from other operating 
  systems, which thus could clearly be argued to not be derived works, and 
  the original limited export table also acted somewhat as a barrier to 
  show a level of distance.

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Larry McVoy]

On Sat, Oct 04, 2003 at 07:52:09PM -0500, Rob Landley wrote:
> On Monday 15 September 2003 00:57, Erik Andersen wrote:
> > On Mon Sep 15, 2003 at 12:17:37AM +0000, Henning P. Schmiedehausen wrote:
> > > Erik Andersen <andersen@codepoet.org> writes:
> > > >When you are done making noise, please explain how a closed
> > > >source binary only product that runs within the context of the
> > > >Linux kernel is not a derivitive work and therefore not subject
> > > >to the terms of the GPL, per the definition given in the kernel
> > > >COPYING file that grants you your limited rights for copying,
> > > >distribution and modification.
> > >
> > > "Because Linus said so".
> >
> > It does not say "Because Linus said so" in the Linux kernel
> > COPYING file, which is the only official document that grants
> > legal permission to copy, distribute and/or modify the kernel.
> 
> Linus clearly and publicly stated his position on binary only kernel modules 
> almost exactly one year ago:

Yeah, but Linus stating his position about a license doesn't mean diddly.
The kernel is licensed under a license, that license is a contract that
people enter into.  To the extent that it is enforceable, that license
determines what happens, Linus can't retroactively decide to interpret
the license a different way.  The license can't enforce things which
the law doesn't allow.  In particular, the law understands a concept of
a boundary.  And Linus' comments notwithstanding, modules are a pretty
clear boundary.  Even the GPL acks this, it knows that anything which
is clearly separable is not covered.
-- 
---
Larry McVoy              lm at bitmover.com          http://www.bitmover.com/lm

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[viro]

On Sat, Oct 04, 2003 at 06:05:21PM -0700, Larry McVoy wrote:
> 
> Yeah, but Linus stating his position about a license doesn't mean diddly.
> The kernel is licensed under a license, that license is a contract that
> people enter into.  To the extent that it is enforceable, that license
> determines what happens, Linus can't retroactively decide to interpret
> the license a different way.  The license can't enforce things which
> the law doesn't allow.  In particular, the law understands a concept of
> a boundary.  And Linus' comments notwithstanding, modules are a pretty
> clear boundary.  Even the GPL acks this, it knows that anything which
> is clearly separable is not covered.

Oh, for fuck sake!  Larry, grep the damn tree for EXPORT_SYMBOL.  And
count them.  _IF_ it would be a relatively sane set of primitives - sure,
no arguments.  It's not.  Nowhere near that.

Conversions from EXPORT_SYMBOL to EXPORT_SYMBOL_GPL are noise.  Why?
Because at any point any exported symbol can disappear.  Period.  For
some of them it's less likely, for some - more, but there was no promise
to preserve that set.  Ever.  Look at them and you will see why - if
we promise to keep all that pile present and working as it used to, we've
got a pitchfork stuck in the kernel guts.

Yes, it would be nice if there was something at least resembling an API.
Get the export list to shrink by 1.5 orders of magnitude and we might
have something to talk about.  That, and get the situation to the point
where additions to the export list would have to be defended - not granted
whenever somebody says "I wanna".  Until then there's no boundary at all.

Right now modules can call _anything_.  Look through the history and you'll
see patches that not only added an export but removed static at the same
chunk.  And you know what?  The guys who would like to pretend that there
is a boundary are the same guys who had destroyed it.  It used to be much
smaller list of exported objects.  Guess who had been pushing for its expansion
until it had lost any semblance of controlled interface?

When additions to interface start happening without any review and without
any percieved need to even explain why you need to add this, this and that -
it stops being an interface.  It's true for any project, not just the kernel.

And I'll bet you anything - if you try to get the damn thing back into shape,
authors of said modules will be out for blood.

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Larry McVoy]

On Sun, Oct 05, 2003 at 03:34:28AM +0100, viro@parcelfarce.linux.theplanet.co.uk wrote:
> On Sat, Oct 04, 2003 at 06:05:21PM -0700, Larry McVoy wrote:
> > 
> > Yeah, but Linus stating his position about a license doesn't mean diddly.
> > The kernel is licensed under a license, that license is a contract that
> > people enter into.  To the extent that it is enforceable, that license
> > determines what happens, Linus can't retroactively decide to interpret
> > the license a different way.  The license can't enforce things which
> > the law doesn't allow.  In particular, the law understands a concept of
> > a boundary.  And Linus' comments notwithstanding, modules are a pretty
> > clear boundary.  Even the GPL acks this, it knows that anything which
> > is clearly separable is not covered.
> 
> Oh, for fuck sake!  Larry, grep the damn tree for EXPORT_SYMBOL.  And
> count them.  _IF_ it would be a relatively sane set of primitives - sure,
> no arguments.  It's not.  Nowhere near that.

You're missing what the law sees as a boundary.  It's really simple,
as far as I can tell, and it doesn't matter how many symbols there are
or are not.  If you can pull out one wad of code and drop in another
and everything works as before then that is a boundary.

A great example of this is a device driver.  Again, I'm not a lawyer
although I've spent a fair amount of time discussing this topic with
lawyers, but it sure seems like that an objective judge would say that
the GPL cannot cross the device driver boundary.

People get all worked up over this but when they do then they should
also claim that system calls are not a boundary either.

By the way, I have no personal or business desire to argue this one way
or the other, I'm not trying to make money off of something like a driver
linked with the kernel or anything remotely similar.  All I'm doing is
telling you what I understand to be the law.  You can do with it what
you will but don't shoot the messenger (or at least don't expect me to
change my tune when you do).
-- 
---
Larry McVoy              lm at bitmover.com          http://www.bitmover.com/lm

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[David Woodhouse]

On Sat, 2003-10-04 at 20:45 -0700, Larry McVoy wrote:
> People get all worked up over this but when they do then they should
> also claim that system calls are not a boundary either.

The first paragraph of the COPYING file makes it entirely clear that
system calls were not considered to be such a boundary.

-- 
dwmw2

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Larry McVoy]

On Sun, Oct 05, 2003 at 11:24:36AM +0100, David Woodhouse wrote:
> On Sat, 2003-10-04 at 20:45 -0700, Larry McVoy wrote:
> > People get all worked up over this but when they do then they should
> > also claim that system calls are not a boundary either.
> 
> The first paragraph of the COPYING file makes it entirely clear that
> system calls were not considered to be such a boundary.

You're forgetting that what the GPL says doesn't matter if it is
unenforceable.  Remember all the people yelling at me that they can
reverse engineer BK in spite of any no-reverse-engineering clauses?
That same logic applies to the GPL, you can't have it both ways.

It doesn't matter what you think, or I think, or Linus thinks.  What
matters is what is legal and what isn't.  
-- 
---
Larry McVoy              lm at bitmover.com          http://www.bitmover.com/lm

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[David Woodhouse]

On Sun, 2003-10-05 at 06:56 -0700, Larry McVoy wrote:
> You're forgetting that what the GPL says doesn't matter if it is
> unenforceable. 

If the licence is invalid then there exists no licence and it is an
offence to use the software in question at _all_.

>  Remember all the people yelling at me that they can
> reverse engineer BK in spite of any no-reverse-engineering clauses?
> That same logic applies to the GPL, you can't have it both ways.

I wasn't doing that yelling.

> It doesn't matter what you think, or I think, or Linus thinks.  What
> matters is what is legal and what isn't. 

Agreed.
 
-- 
dwmw2

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[David Woodhouse]

On Sat, 2003-10-04 at 18:05 -0700, Larry McVoy wrote:
> Yeah, but Linus stating his position about a license doesn't mean diddly.
> The kernel is licensed under a license, that license is a contract that
> people enter into.  To the extent that it is enforceable, that license
> determines what happens, Linus can't retroactively decide to interpret
> the license a different way.  The license can't enforce things which
> the law doesn't allow.  In particular, the law understands a concept of
> a boundary. 

I agree. Linus' comments on the matter, except of course his original
exception for userspace which has been there since 1993, are an
irrelevant _interpretation_.

All that is relevant is the meaning of the original licence.

>  And Linus' comments notwithstanding, modules are a pretty
> clear boundary.  Even the GPL acks this, it knows that anything which
> is clearly separable is not covered.

This statement is in conflict with the presence of the exception for
userspace which precedes the text of the GPL in the Linux COPYING file.

The presence of the exception makes it clear that, without such
exception, userspace would have been considered to be a derived work in
the terminology of the original licence. Otherwise, the exception would
of course have been redundant.

If userspace would be considered a derived work without explicit
exception, then so are kernel modules. They have no such explicit
exception, and are hence not permitted.

I'll grant you that Linus' witterings in public on the matter would
probably prevent him _personally_ from bringing suit against a
distributor of binary-only modules, on the principle of equitable
estoppel.

That doesn't stop me or anyone else from doing so though.

-- 
dwmw2

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[David Lang]

On Sun, 5 Oct 2003, David Woodhouse wrote:

> The presence of the exception makes it clear that, without such
> exception, userspace would have been considered to be a derived work in
> the terminology of the original licence. Otherwise, the exception would
> of course have been redundant.
>
> If userspace would be considered a derived work without explicit
> exception, then so are kernel modules. They have no such explicit
> exception, and are hence not permitted.

no, all the presence of the userspace exception means is that someone
attempted to make the claim that you would only be allowed to run GPL
software on a GPL kernel and Linus wanted to make it absolutly clear that
that wasn't the case.

trying to claim otherwise, even without a specific 'userspace exception'
is along the same lines as what SCO is doing, (anything that ever ran on
the same box as Sys V is part of SysV)

why do people realize how stupid this argument when SCO makes it, but
somehow when it's made on behalf of the GPL it somehow seems sane?

as got the GPL_only stuff, I am seriously worried about people defining
something and then declaring that anything that uses it in any way must be
a derived work, that there is no other legitimate way to use it. how many
people would buy this argument if it was being made about some function in
a piece of hardware? (i.e., if you use this function on this 802.11 card
then your software is obviously a derivitive of our driver so we get all
the rights to it)

David Lang

-- 
"Debugging is twice as hard as writing the code in the first place.
Therefore, if you write the code as cleverly as possible, you are,
by definition, not smart enough to debug it." - Brian W. Kernighan

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[David Woodhouse]

On Sun, 2003-10-05 at 04:32 -0700, David Lang wrote:
> why do people realize how stupid this argument when SCO makes it, but
> somehow when it's made on behalf of the GPL it somehow seems sane?

The distinction to be drawn here is between that which is allowed for by
copyright law, and that which is not.

Let us briefly assume, for the sake of argument, that the Linux kernel
was released under the Creosote Public Licence, a licence which requires
each licensee to perform a daily ritual of bathing in creosote, and to
release _all_ future work, even unrelated work, of his own under the
same licence.

If you don't like the terms of the CPL, you have the option of not using
the work in question.

If you use the work, however, you are bound by those terms. 

Let us assume, also merely for the sake of argument, that you can
produce a kernel module which is not so closely tied to the kernel that
it would be considered a derivative work in copyright law.

However, the CPL under which the kernel is released still requires that
you release this work under the CPL, even though under copyright law it
is not a derivative work.

You still have the same choice you had before. You may accept the
licence of the Linux kernel, continue to perform your daily bathing in
creosote and release your new work under the appropriate licence -- or
you may decline the licence of the Linux kernel and refrain from using
it at all.

If you take the latter option, then you may perform no testing on your
binary module, since that would require use of the Linux kernel. You may
use the Linux kernel nowhere within your organisation.

In this situation, it's not just about your module itself being an
infringing copy of a copyright work, but your copy of the Linux kernel
_itself_ being an infringing copy.

Obviously the GNU General Public License makes no mention of creosote,
and its use with the Linux kernel does not require that you license
_all_ future work under the GPL. 

I have made a statement about my opinion of what the licence of the
Linux kernel _does_ in fact require. It is my belief that the author of
the GNU General Public License is in agreement with me. 

In particular, in this case, I think the presence of the userspace
exception makes it _very_ clear that the meaning of 'derived work' would
otherwise have included both userspace and loadable modules. 

Obviously this is neither true or false until the matter is settled in a
court; it's all conjecture. But I think a court will agree with me.

>  how many people would buy this argument if it was being made about some 
> function in a piece of hardware? (i.e., if you use this function on this 
> 802.11 card then your software is obviously a derivitive[sic] of our driver 
> so we get all the rights to it)

I'd buy the _argument_, even if the licence were to require the ritual
sacrifice of my first-born child. I wouldn't buy the card in question
and use that function though. I have that choice.

If I _did_ choose to use the function in question, I would not then
whinge that it's not fair when they come for my first-born. That was the
agreement I entered into, after all.

-- 
dwmw2

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Andre Hedrick]

You are so young and fresh to the game, it is cute.

http://www.gcom.com/home/support/whitepapers/linux-gnu-license.html

Try doing a little more checking of your history Rob.

The position was set in 1995 and to many who objected let it unchallanged.
Now regardless if you knew of this information or not when you entered
into kernel development, tough!  I did not know about it and have to suck
it up that binary modules are permitted, period.  So if other people can
make money, then I will too, and encourage others to persue also.

If you want to try and strike with the iron is cold bring a lunch.
Otherwise bring your lawyers when I publish freed_symbols and will eat
your lame position for a snack and spit it out because of bad taste.

Tell me I can not publish a GPL w/ source code project which returns the
original API's to their normal place in history, and I will show you that
I can still draw the string on a bow.

Tell anyone they can not change anything they want in the kernel, and you
impose a restriction and you lose your right to use by the terms of the
license called GPL.  Funny how the other side of the sword is ignored
even when it is splitting your forehead wide open.

So have a nice day.

Andre Hedrick
LAD Storage Consulting Group

On Sat, 4 Oct 2003, Rob Landley wrote:

> On Monday 15 September 2003 00:57, Erik Andersen wrote:
> > On Mon Sep 15, 2003 at 12:17:37AM +0000, Henning P. Schmiedehausen wrote:
> > > Erik Andersen <andersen@codepoet.org> writes:
> > > >When you are done making noise, please explain how a closed
> > > >source binary only product that runs within the context of the
> > > >Linux kernel is not a derivitive work and therefore not subject
> > > >to the terms of the GPL, per the definition given in the kernel
> > > >COPYING file that grants you your limited rights for copying,
> > > >distribution and modification.
> > >
> > > "Because Linus said so".
> >
> > It does not say "Because Linus said so" in the Linux kernel
> > COPYING file, which is the only official document that grants
> > legal permission to copy, distribute and/or modify the kernel.
> 
> Linus clearly and publicly stated his position on binary only kernel modules 
> almost exactly one year ago:
> 
> http://groups.google.com/groups?selm=Pine.LNX.4.44.0210170958340.6739-100000%40home.transmeta.com.lucky.linux.kernel
> 
> He basically said there IS no module exception to the GPL, it's just a 
> question of what is and is not a derived work.
> 
> The kernel developers have marked up portions of the API to indicate "we 
> consider anything that needs to access this deeply internal bit to be a 
> derived work, hence subject to the GPL".  That's what GPL_ONLY _means_.  
> Needing to re-export that therefore opens you up to a lawsuit.  (Whether you 
> can defend yourself in court from that lawsuit is always an open question, 
> but by adding GPL_ONLY markup the developers made their intent much more 
> clear, which is unlikely to help you convince a judge of your interpretation 
> if you explicitly undo that markup and then claim the license doesn't apply 
> to you...)
> 
> Here's the relevant section of the above posting from Linus:
> 
> -----
> 
> I will re-iterate my stance on the GPL and kernel modules:
> 
>   There is NOTHING in the kernel license that allows modules to be 
>   non-GPL'd. 
> 
>   The _only_ thing that allows for non-GPL modules is copyright law, and 
>   in particular the "derived work" issue. A vendor who distributes non-GPL 
>   modules is _not_ protected by the module interface per se, and should 
>   feel very confident that they can show in a court of law that the code 
>   is not derived.
> 
>   The module interface has NEVER been documented or meant to be a GPL 
>   barrier. The COPYING clearly states that the system call layer is such a 
>   barrier, so if you do your work in user land you're not in any way 
>   beholden to the GPL. The module interfaces are not system calls: there 
>   are system calls used to _install_ them, but the actual interfaces are
>   not.
> 
>   The original binary-only modules were for things that were pre-existing 
>   works of code, ie drivers and filesystems ported from other operating 
>   systems, which thus could clearly be argued to not be derived works, and 
>   the original limited export table also acted somewhat as a barrier to 
>   show a level of distance.
> 

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[viro]

On Sat, Oct 04, 2003 at 11:40:22PM -0700, Andre Hedrick wrote:
> 
> Tell me I can not publish a GPL w/ source code project which returns the
> original API's to their normal place in history, and I will show you that
> I can still draw the string on a bow.

_What_ original API?  I agree that silent adding _GPL to existing symbol
is obnoxious and warrants a patch that would revert the change.

However, if tomorrow the exported function disappears completely - tough
luck.  Nobody had ever promised to keep this "API" unchanged.  It's not
that it had been changed just for kicks (after all, you get to do changes
in a bunch of in-tree drivers are such change), but such changes had happened
and will happen.  And there's nothing you can do about that.

And folks, let's be honest.  Sturgeon was an optimist.  Way more than 90%
of code is crap.  The only way around that is to have a bunch of creatively
sadistic bastards go through said code and rip the authors a new one for
every hole they find (and yes, that includes ripping new ones to each other).

Judging by the vendor drivers that doesn't happen.  I don't care why that
doesn't happen - be it "they'll buy it anyway" or "we have no resources"
or "it's rude to the people who had done the original work" or "what do
you mean, review?".  Whatever.  Unless I have very good reasons to believe
that particular piece of code had been done right, crap it is.  Plain and
simple statistics.

Code from unknown programmers presumably written to unknown specifications
that had presumably passed unknown QA by unknown reviewers and testers with
unknown results and then had been shipped with unknown amount of pressure
exerted by sales?  Geez...  What a wonderful reason to assume that it would
be better than average...

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[David Woodhouse]

On Sat, 2003-10-04 at 23:40 -0700, Andre Hedrick wrote:
> You are so young and fresh to the game, it is cute.
> 
> http://www.gcom.com/home/support/whitepapers/linux-gnu-license.html

That's quite amusingly erroneous. The idea that you could use the Linux
kernel having stripped out all code belonging to Alan so that his
opinion is not relevant is one that I find frankly hilarious.

This is not about your work just being a derived work of Alan's under
copyright law. This is about you not complying to the licence under
which Alan's code is released and hence it is _his_ code, including I
believe a significant part of the networking code, which is an
'infringing copy of a copyright work' in law; since you use it without
licence.

-- 
dwmw2

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Andre Hedrick]

David,

It is about the fact that Linus on his own set a position that modules are
permitted.  Now if you want to take the position that one can not modify
and redistribute the modified kernel in source, you are imposing a
restriction.  One is free to add or subtract content and redistribute.
Combine this with the position that modules are permitted regardless and
there is no copyright issue.

If I wanted to be rude, I could take the changes I made and copyright the
combined work and make it so others could not use that version without
permission.  This is exactly what the large distributions do with their
product.  Surprized, don't be it is all a stupid game of no you can't and
yes I can.

So when should I expect your letter ?

How will you phrase the content given "freed_symbols" will be GPL ?

How will you attack any vendor using it ?

As the author of the GPL code, I will show that I can still pull a bow
string.  If I use it for whatever reason or combine it into any kernel I
distribute as source where are your issues?  You have none.

This is all a game of who is peeing in whose soup.

Neither you or I can do anything, provide the vendor who is using Linux
publishes their source fork.  I also dare you to stop them, because you
can't.

Cheers, and keep up the joust!

Andre


Andre Hedrick
LAD Storage Consulting Group

On Sun, 5 Oct 2003, David Woodhouse wrote:

> On Sat, 2003-10-04 at 23:40 -0700, Andre Hedrick wrote:
> > You are so young and fresh to the game, it is cute.
> > 
> > http://www.gcom.com/home/support/whitepapers/linux-gnu-license.html
> 
> That's quite amusingly erroneous. The idea that you could use the Linux
> kernel having stripped out all code belonging to Alan so that his
> opinion is not relevant is one that I find frankly hilarious.
> 
> This is not about your work just being a derived work of Alan's under
> copyright law. This is about you not complying to the licence under
> which Alan's code is released and hence it is _his_ code, including I
> believe a significant part of the networking code, which is an
> 'infringing copy of a copyright work' in law; since you use it without
> licence.
> 
> -- 
> dwmw2
> 
> 

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[David Woodhouse]

On Sun, 2003-10-05 at 12:21 -0700, Andre Hedrick wrote:
> David,
> 
> It is about the fact that Linus on his own set a position that modules are
> permitted. 

It is indeed. And about that fact that by the time he made that
declaration, he was not in a position to make it unilaterally.


>  Now if you want to take the position that one can not modify
> and redistribute the modified kernel in source, you are imposing a
> restriction.

A restriction in this case which I assert was present in the original
licence; not an 'additional restriction'.

If I wanted to distribute my code without restrictions, I'd have
contributed to a BSD kernel, or released my code under the terms 
"GPL but not if you really really don't fancy it".

I don't. I didn't.

> If I wanted to be rude, I could take the changes I made and copyright the
> combined work and make it so others could not use that version without
> permission. 

Not without being in violation of the original licence. This isn't about
the resulting combined work, but about your permission to use the
_original_.

> Neither you or I can do anything, provide the vendor who is using Linux
> publishes their source fork.  I also dare you to stop them, because you
> can't.

Believe me, there are ways this can be achieved.
 
-- 
dwmw2

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Andre Hedrick]

On Sun, 5 Oct 2003, David Woodhouse wrote:

> On Sun, 2003-10-05 at 12:21 -0700, Andre Hedrick wrote:
> > David,
> > 
> > It is about the fact that Linus on his own set a position that modules are
> > permitted. 
> 
> It is indeed. And about that fact that by the time he made that
> declaration, he was not in a position to make it unilaterally.

Regardless, nobody stopped him at that time and thus a right of way has
been granted and can not be revoked.

> 
> >  Now if you want to take the position that one can not modify
> > and redistribute the modified kernel in source, you are imposing a
> > restriction.
> 
> A restriction in this case which I assert was present in the original
> licence; not an 'additional restriction'.

See above, the boss changed the rules and nobody challanged it.
It sticks like glue and we all have to suck up the point and move forward.

> If I wanted to distribute my code without restrictions, I'd have
> contributed to a BSD kernel, or released my code under the terms 
> "GPL but not if you really really don't fancy it".
> 
> I don't. I didn't.
> 
> > If I wanted to be rude, I could take the changes I made and copyright the
> > combined work and make it so others could not use that version without
> > permission. 
> 
> Not without being in violation of the original licence. This isn't about
> the resulting combined work, but about your permission to use the
> _original_.
> 
> > Neither you or I can do anything, provide the vendor who is using Linux
> > publishes their source fork.  I also dare you to stop them, because you
> > can't.
> 
> Believe me, there are ways this can be achieved.

Sure, and I am out there proving the points that will undermine your
arguements.

Whatever happened to "World Domination" according to TUX ?
Whatever happened to 'having a choice' ?

Lets assume you are correct, and the effect is a "Tar Baby".

Your claims that anything which loads into a kernel is automatically a
derived work.  Thus the effect of an original work loading into a gpl work
force the original work to be GPL.  This is a joke and will never see a
second in any court.

Cheers,

Andre

For those who do not know about history and english archers, go read.

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[David Woodhouse]

On Sun, 2003-10-05 at 13:14 -0700, Andre Hedrick wrote:
> Regardless, nobody stopped him at that time and thus a right of way has
> been granted and can not be revoked.

Wrong. While Linus' statement does bar him from personally suing you, it
doesn't stop anyone else.

> See above, the boss changed the rules and nobody challanged it.

He did not have authority, by that time, to change the rules. Neither
was he unchallenged.

> Whatever happened to "World Domination" according to TUX ?
> Whatever happened to 'having a choice' ?

You do have a choice. You can use software under the terms of the
licence under which it's released, or you can choose not to use it.

> Lets assume you are correct, and the effect is a "Tar Baby".
> 
> Your claims that anything which loads into a kernel is automatically a
> derived work.  Thus the effect of an original work loading into a gpl work
> force the original work to be GPL.  This is a joke and will never see a
> second in any court.

You really aren't paying attention, are you? Your copyright (or lack of
it as a derivative work) on your own module is largely irrelevant to my
argument.

The GPL says you may use the kernel _itself_ but only with certain
restrictions.

My claim is that the GPL forbids you from loading a non-GPL'd module.
Not that if you do so, the non-GPL'd module becomes a derived work, but
that in doing do you are violating the licence under which you received
the _kernel_ and hence you must immediately cease using the _kernel_.

Just like if the GPL required you to bathe in creosote daily and one day
you forgot.

I repeat, for the hard of understanding:

I am not asserting that if you manage to produce a loadable module which
a court would rule is not a derivative work, you would not be allowed to
distribute that.

I am asserting that if you do so, you are disobeying the restrictions on
your use of the kernel itself, and hence you would not be able to use
the kernel. You could use your own module, but not the Linux kernel.

-- 
dwmw2

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Andre Hedrick]

Will give you your point, regardless that is is wrong.

If loading a NON-GPL module is in violation of the license then enforce
the license to prevent one from loading.

Legally you can not, thus the license for which you claim is in violation,
violates itself.  Restriction of usage.

Well the right-of-way is now there little chance of it being taken back.

GPL sucks as a general license, yet it was the best thing going at the
time.  It has a means to prevent us from moving forward to migrate to a
better license.

Cheers,

Andre Hedrick
LAD Storage Consulting Group

On Sun, 5 Oct 2003, David Woodhouse wrote:

> On Sun, 2003-10-05 at 13:14 -0700, Andre Hedrick wrote:
> > Regardless, nobody stopped him at that time and thus a right of way has
> > been granted and can not be revoked.
> 
> Wrong. While Linus' statement does bar him from personally suing you, it
> doesn't stop anyone else.
> 
> > See above, the boss changed the rules and nobody challanged it.
> 
> He did not have authority, by that time, to change the rules. Neither
> was he unchallenged.
> 
> > Whatever happened to "World Domination" according to TUX ?
> > Whatever happened to 'having a choice' ?
> 
> You do have a choice. You can use software under the terms of the
> licence under which it's released, or you can choose not to use it.
> 
> > Lets assume you are correct, and the effect is a "Tar Baby".
> > 
> > Your claims that anything which loads into a kernel is automatically a
> > derived work.  Thus the effect of an original work loading into a gpl work
> > force the original work to be GPL.  This is a joke and will never see a
> > second in any court.
> 
> You really aren't paying attention, are you? Your copyright (or lack of
> it as a derivative work) on your own module is largely irrelevant to my
> argument.
> 
> The GPL says you may use the kernel _itself_ but only with certain
> restrictions.
> 
> My claim is that the GPL forbids you from loading a non-GPL'd module.
> Not that if you do so, the non-GPL'd module becomes a derived work, but
> that in doing do you are violating the licence under which you received
> the _kernel_ and hence you must immediately cease using the _kernel_.
> 
> Just like if the GPL required you to bathe in creosote daily and one day
> you forgot.
> 
> I repeat, for the hard of understanding:
> 
> I am not asserting that if you manage to produce a loadable module which
> a court would rule is not a derivative work, you would not be allowed to
> distribute that.
> 
> I am asserting that if you do so, you are disobeying the restrictions on
> your use of the kernel itself, and hence you would not be able to use
> the kernel. You could use your own module, but not the Linux kernel.
> 
> -- 
> dwmw2
> 
> 

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Larry McVoy]

On Sun, Oct 05, 2003 at 09:34:40PM +0100, David Woodhouse wrote:
> The GPL says you may use the kernel _itself_ but only with certain
> restrictions.
> 
> My claim is that the GPL forbids you from loading a non-GPL'd module.
> Not that if you do so, the non-GPL'd module becomes a derived work, but
> that in doing do you are violating the licence under which you received
> the _kernel_ and hence you must immediately cease using the _kernel_.

Your claim is not, as far as I know, supported by the law or the GPL
itself.

GPL v2 section 2:
    "These requirements apply to the modified work as a whole.
    If identifiable sections of that work are not derived from the
    Program, and can be reasonably considered independent and separate
    works in themselves, then this License, and its terms, do not apply"

That leaves the question of what "can be reasonably considered independent
and separate works".  I've both heard about other companies researching
this and I've done it myself.  The lawyers came to the same conclusion,
independently.  In software, what constitutes an independent work is
something which can be pulled out and have another implementation dropped
in and the rest of the system can't tell the difference.

A very obvious boundary is user vs kernel, nobody here thinks that because
some application runs on a GPLed kernel that application is GPLed.
Some people here may _pretend_ they think that so that they can argue
that Linus made an "exception" for user land applications but that's
just self serving posturing and I'm sure those people know that (just
as I'm sure there will be 50 flaming replies saying that is not at all
what they think.  Politicians are the same everywhere).

Another boundary is a tarball.  If it weren't for the above clause of
the GPL then anything combined in a tarball with a GPLed work would be
considered GPLed.  Even RMS knew that wouldn't fly.

A less obvious boundary, and the one that got me into this, is the storage
of a GPLed source file in a source management system.  Does that mean that
the metadata used to store that file is GPLed?  At one point I was worried
about this (why?  Damn good question, in retrospect it is a "don't care",
I didn't create the metadata so I don't own it anyway so why do I care if
it is GPLed or not?  Whatever, at one point I cared).  I spent more than
a lot of you make in a year in legal fees looking into it and that's where 
I learned about boundaries.  The law has pretty clear ideas about boundaries
and it doesn't matter what you think or I think or the GPL thinks, the
boundaries are there and the GPL can't cross them.  The conclusion of the
lawyers was that no, putting a GPLed file into an SCM system in no way 
makes the SCM metadata GPLed.  BTW, I asked RMS about this and he of course
refused to accept that, his position is that the metadata would be GPLed,
nice to see he is consistent :)

A much more obvious example than the SCM one is a device driver or a module.
That's so cut and dried it isn't even open to debate in the eyes of the 
law.  It's a hard and fast boundary, the GPL can't cross it no matter what
people think or want (on either side).

That's why I think that your claim is not supported, by the GPL or
(far more importantly) the law.  While I'm no lawyer I'm perhaps more
qualified than some people on this list since I've actually spent a pile
of money researching this.  I'm sure that someone with more money could
buy^H^H^Hpay some lawyers try and make an opposing view stick but I'm
equally sure that those of you without money don't have an iceballs'
chance in hell of making an opposing view stick.  Talk is cheap, legal
decisions are expensive.

Once again, please note that I don't make money off the kernel or any other
GPLed product (we ship diff & patch with BK but we also provide source for
all our changes, they aren't substantial nor are they money makers).  So I
have no vested interest in which way this works out.  I'm simply passing on
what I've learned, I'm more or less one of you who has actually spent a lot
of money getting legal opinions on the topic.
-- 
---
Larry McVoy              lm at bitmover.com          http://www.bitmover.com/lm

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[David Lang]

not to disagree with Larry in what he posted below, but there is nothing
in what he says that at all means that if you paste in code from one side
of a barrier to the other side the result doesn't need to be GPL'd

so the barriers can be delibratly bridged, but if you are careful about
your code on one side of the barrier you don't have to worry about what's
on the other side.

David Lang

On Sun, 5 Oct 2003, Larry McVoy wrote:

> Date: Sun, 5 Oct 2003 18:22:12 -0700
> From: Larry McVoy <lm@bitmover.com>
> To: David Woodhouse <dwmw2@infradead.org>
> Cc: Andre Hedrick <andre@linux-ide.org>, Rob Landley <rob@landley.net>,
>      Henning P. Schmiedehausen <hps@intermeta.de>,
>      linux-kernel@vger.kernel.org
> Subject: Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]
>
> On Sun, Oct 05, 2003 at 09:34:40PM +0100, David Woodhouse wrote:
> > The GPL says you may use the kernel _itself_ but only with certain
> > restrictions.
> >
> > My claim is that the GPL forbids you from loading a non-GPL'd module.
> > Not that if you do so, the non-GPL'd module becomes a derived work, but
> > that in doing do you are violating the licence under which you received
> > the _kernel_ and hence you must immediately cease using the _kernel_.
>
> Your claim is not, as far as I know, supported by the law or the GPL
> itself.
>
> GPL v2 section 2:
>     "These requirements apply to the modified work as a whole.
>     If identifiable sections of that work are not derived from the
>     Program, and can be reasonably considered independent and separate
>     works in themselves, then this License, and its terms, do not apply"
>
> That leaves the question of what "can be reasonably considered independent
> and separate works".  I've both heard about other companies researching
> this and I've done it myself.  The lawyers came to the same conclusion,
> independently.  In software, what constitutes an independent work is
> something which can be pulled out and have another implementation dropped
> in and the rest of the system can't tell the difference.
>
> A very obvious boundary is user vs kernel, nobody here thinks that because
> some application runs on a GPLed kernel that application is GPLed.
> Some people here may _pretend_ they think that so that they can argue
> that Linus made an "exception" for user land applications but that's
> just self serving posturing and I'm sure those people know that (just
> as I'm sure there will be 50 flaming replies saying that is not at all
> what they think.  Politicians are the same everywhere).
>
> Another boundary is a tarball.  If it weren't for the above clause of
> the GPL then anything combined in a tarball with a GPLed work would be
> considered GPLed.  Even RMS knew that wouldn't fly.
>
> A less obvious boundary, and the one that got me into this, is the storage
> of a GPLed source file in a source management system.  Does that mean that
> the metadata used to store that file is GPLed?  At one point I was worried
> about this (why?  Damn good question, in retrospect it is a "don't care",
> I didn't create the metadata so I don't own it anyway so why do I care if
> it is GPLed or not?  Whatever, at one point I cared).  I spent more than
> a lot of you make in a year in legal fees looking into it and that's where
> I learned about boundaries.  The law has pretty clear ideas about boundaries
> and it doesn't matter what you think or I think or the GPL thinks, the
> boundaries are there and the GPL can't cross them.  The conclusion of the
> lawyers was that no, putting a GPLed file into an SCM system in no way
> makes the SCM metadata GPLed.  BTW, I asked RMS about this and he of course
> refused to accept that, his position is that the metadata would be GPLed,
> nice to see he is consistent :)
>
> A much more obvious example than the SCM one is a device driver or a module.
> That's so cut and dried it isn't even open to debate in the eyes of the
> law.  It's a hard and fast boundary, the GPL can't cross it no matter what
> people think or want (on either side).
>
> That's why I think that your claim is not supported, by the GPL or
> (far more importantly) the law.  While I'm no lawyer I'm perhaps more
> qualified than some people on this list since I've actually spent a pile
> of money researching this.  I'm sure that someone with more money could
> buy^H^H^Hpay some lawyers try and make an opposing view stick but I'm
> equally sure that those of you without money don't have an iceballs'
> chance in hell of making an opposing view stick.  Talk is cheap, legal
> decisions are expensive.
>
> Once again, please note that I don't make money off the kernel or any other
> GPLed product (we ship diff & patch with BK but we also provide source for
> all our changes, they aren't substantial nor are they money makers).  So I
> have no vested interest in which way this works out.  I'm simply passing on
> what I've learned, I'm more or less one of you who has actually spent a lot
> of money getting legal opinions on the topic.
> --
> ---
> Larry McVoy              lm at bitmover.com          http://www.bitmover.com/lm
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
>

-- 
"Debugging is twice as hard as writing the code in the first place.
Therefore, if you write the code as cleverly as possible, you are,
by definition, not smart enough to debug it." - Brian W. Kernighan

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Larry McVoy]

On Sun, Oct 05, 2003 at 06:37:18PM -0700, David Lang wrote:
> not to disagree with Larry in what he posted below, but there is nothing
> in what he says that at all means that if you paste in code from one side
> of a barrier to the other side the result doesn't need to be GPL'd

Agreed.  If you are making a derived work, and pasting in stuff from a
GPLed source, your work is definitely derived and you are GPLed.  And I'd
be with everyone else ripping you a new one if you didn't GPL it.

I hope people understand that I respect the GPL and what it has accomplished.
It's a great thing.  People should not abuse it.  My comments which may seem
against that are legal in their basis, not my personal goals.
-- 
---
Larry McVoy              lm at bitmover.com          http://www.bitmover.com/lm

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Maciej Zenczykowski]

> You are so young and fresh to the game, it is cute.
> 
> http://www.gcom.com/home/support/whitepapers/linux-gnu-license.html

Can a module even be considered LGPL?  After all a module interfaces with
the kernel via including files from the kernel source - doesn't this
automatically mean that it is a derived work of at least a few of the
kernel headers (the module specific ones for example).  These headers
contribute code to the module as well: INC_MOD_USE_COUNT and the like...
And since the kernel is GPLed doesn't this mean that the entire module is
GPLed?

On the other hand any running program on linux dynamically links (via 
syscalls) against the kernel... I think everyone agrees that dynamically 
linking against the kernel in this manner should be allowed and not a 
violation of the GPL of the kernel source...

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Andre Hedrick]

Usage of the kernel headers only, is using the effective API for that
snapshot.  It is clear in many copyright cases under the terms of fair use
and reverse engineering to obtain operational functionality is legal and
upheld.

Touching or using copied C-code is fatal.

Reading the C-code and creating another work which is identical in
functionality and completely original, it is not derived regardless what
anybody thinks or tells you.  GPL only protects the actual file or
document.  It is a total waste on protecting the content expressed.

People confuse the two, and must admit I did so in the past.

Cheers,

Andre Hedrick
LAD Storage Consulting Group

On Sun, 5 Oct 2003, Maciej Zenczykowski wrote:

> > You are so young and fresh to the game, it is cute.
> > 
> > http://www.gcom.com/home/support/whitepapers/linux-gnu-license.html
> 
> Can a module even be considered LGPL?  After all a module interfaces with
> the kernel via including files from the kernel source - doesn't this
> automatically mean that it is a derived work of at least a few of the
> kernel headers (the module specific ones for example).  These headers
> contribute code to the module as well: INC_MOD_USE_COUNT and the like...
> And since the kernel is GPLed doesn't this mean that the entire module is
> GPLed?
> 
> On the other hand any running program on linux dynamically links (via 
> syscalls) against the kernel... I think everyone agrees that dynamically 
> linking against the kernel in this manner should be allowed and not a 
> violation of the GPL of the kernel source...
> 
> 
> -
> To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
> the body of a message to majordomo@vger.kernel.org
> More majordomo info at  http://vger.kernel.org/majordomo-info.html
> Please read the FAQ at  http://www.tux.org/lkml/
> 

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[David Woodhouse]

On Sun, 2003-10-05 at 12:47 -0700, Andre Hedrick wrote:
> Reading the C-code and creating another work which is identical in
> functionality and completely original, it is not derived regardless what
> anybody thinks or tells you.  GPL only protects the actual file or
> document. 

...and continues to do so. So if the licence of the original says you
may use it only if you do not create such a work of your own, then as
soon as you create your own, you lose your right to continue to use the
original.

-- 
dwmw2

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Andre Hedrick]

Use maybe but read nope.

Now you have crossed the line in GPL, where you have restricted usaged.

You lose :-P

Cheers,

Andre Hedrick
LAD Storage Consulting Group

On Sun, 5 Oct 2003, David Woodhouse wrote:

> On Sun, 2003-10-05 at 12:47 -0700, Andre Hedrick wrote:
> > Reading the C-code and creating another work which is identical in
> > functionality and completely original, it is not derived regardless what
> > anybody thinks or tells you.  GPL only protects the actual file or
> > document. 
> 
> ...and continues to do so. So if the licence of the original says you
> may use it only if you do not create such a work of your own, then as
> soon as you create your own, you lose your right to continue to use the
> original.
> 
> -- 
> dwmw2
> 
> 

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Arjan van de Ven]

[-- Attachment #1: Type: text/plain, Size: 613 bytes --]

On Sun, 2003-10-05 at 21:32, Maciej Zenczykowski wrote:
> 
> On the other hand any running program on linux dynamically links (via 
> syscalls) against the kernel... I think everyone agrees that dynamically 
> linking against the kernel in this manner should be allowed and not a 
> violation of the GPL of the kernel source...

"linking" is a bit tricky here. Traditionally linking involves resolving
addresses of symbols from the other part; system calls don't have any of
this, they are hard coded, documented and fixed numbers... numbers that
also work on BSD and AIX (in the linux personality).


[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Valdis.Kletnieks]

[-- Attachment #1: Type: text/plain, Size: 1284 bytes --]

On Sun, 14 Sep 2003 00:10:27 PDT, Andre Hedrick said:

> Nope, you made threats.

On Sat, 13 Sep 2003, Erik Andersen wrote:
> I'll go even farther, and say that one might call the GPL_ONLY
> symbols an "effective technological measure" that "effectively
> controls access to a work" and "effectively protects a right of a

to which you replied:

On Sat, 13 Sep 2003 19:40:52 -0700, Andre Hedrick said:
> Go have your "DADDY" write another legal letter for you and send it my
> way.  I will be happy to shove it down your pie hole.

Erik postulates a "one MIGHT" legal argument, and gets threatened with bodily
harm?  I think you've given up any moral high ground regarding threats.

You may want to just take a few days off - you're apparently not attached to
the same reality as the rest of us:

> If one reads ./include/linux/module.h
>
> It clearly states any license is acceptable.

Maybe if you apply the Bible Code to it, it's clearly stated, but all I see is
a reference that the MODULE_LICENSE macro will accept a parameter of
"Proprietary", and then goes on to say "it's there so the module in question
can be treated properly - just like a Jew in Warsaw in 1941 had to wear a star".

There. I said it.  The esteemed Mr Godwin says we're now free to get on with our lives.



[-- Attachment #2: Type: application/pgp-signature, Size: 226 bytes --]

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Henning P. Schmiedehausen]

Andre Hedrick <andre@linux-ide.org> writes:


>It is coming and the intent is to return all the stolen symbols.
>It is free for anyone to use and enjoy the usage of Linux once again.
>So everyone get in line and SUE ME for GPL'ed drivers.

[... module code that would re-export GPL-marked symbols as non-GPL-marked snipped ...]

Well,

generally speaking, you're of course right. You're simply using the
loophole of Linus' agreement to binary only modules to use a fully
GPL'ed module (which might use the _GPL symbols), then consider the
aggregation to be under GPL (IMHO correct) and then consider this
aggregation of kernel and your module to be still covered by Linus'
agreement (don't know whether this is true. You might want to actually
ask Linus himself... ;-) )

IMHO doing so might be the best way to make Linus (as the main
copyright holder on the kernel source) to simply revoke the "I won't
object to loading binary only modules in the GPL'ed kernel" agreement
and simply say "From Kernel 2.6 on, every aggregation of modules in
kernel space is considered to be an aggregation in the GPL v2 sense of
meaning as covered by the GPL v2. So if you want to load a module,
it's code is better be GPL'ed too".

In other words: You might force the copyright holder(s) of the Linux
kernel to kill your business model dead.

Is this really what you want? It's basically the same thing that
people do with BitKeeper and Mr. McVoy: Annoy the people that you
might depend on long enough and they might stop being friendly to you
[1]. You might want to ask yourself if this _really_ is what you want
to achieve.

	... just my random 0,02 Euro-Cent
		Henning

[1] The people opposing to BK might not use it themselves but they n
might be heavily using a project which might not be able to go on
without BK because the main developers have stated often enough, that
they won't be able to cope with the work load without BK: The Linux
kernel.
-- 
Dipl.-Inf. (Univ.) Henning P. Schmiedehausen          INTERMETA GmbH
hps@intermeta.de        +49 9131 50 654 0   http://www.intermeta.de/

Java, perl, Solaris, Linux, xSP Consulting, Web Services 
freelance consultant -- Jakarta Turbine Development  -- hero for hire

"Dominate!! Dominate!! Eat your young and aggregate! I have grotty silicon!" 
      -- AOL CD when played backwards  (User Friendly - 200-10-15)

Re: freed_symbols [Re: People, not GPL [was: Re: Driver Model]]

[Justin Cormack]

On Mon, 2003-09-15 at 01:16, Henning P. Schmiedehausen wrote:
> Andre Hedrick <andre@linux-ide.org> writes:
> 
> 
> >It is coming and the intent is to return all the stolen symbols.
> >It is free for anyone to use and enjoy the usage of Linux once again.
> >So everyone get in line and SUE ME for GPL'ed drivers.
> 
> [... module code that would re-export GPL-marked symbols as non-GPL-marked snipped ...]
> 
> Well,
> 
> generally speaking, you're of course right. You're simply using the
> loophole of Linus' agreement to binary only modules to use a fully
> GPL'ed module (which might use the _GPL symbols), then consider the
> aggregation to be under GPL (IMHO correct) and then consider this
> aggregation of kernel and your module to be still covered by Linus'
> agreement (don't know whether this is true. You might want to actually
> ask Linus himself... ;-) )

actually (not that I was following the thread too closely) I thought the
GPL point in the dispute came down to the fact that as the kernel is
under GPL, you can change the export_symbols anywhere you like (under
the GPL) so you can export anything to a binary module.

To really enforce export to GPL only would be simple, you have the file
that exports symbols not under the GPL (and probably the loader to
really enforce it). Writing the license for that file/file loader
combination would be hard but maybe doable...