[Madwifi-devel] Re: madwifi licensing vs kernel (was: netgear wg511t - wireless network card)

[Axel T. <Axel.Thimm@ATrpms.net>]

On Sat, Feb 18, 2006 at 02:08:26PM +0100, Arjan van de Ven wrote:
> On Sat, 2006-02-18 at 13:55 +0100, Axel Thimm wrote:
> > On Sat, Feb 18, 2006 at 12:37:15PM +0100, Arjan van de Ven wrote:
> > > well that and the fact that many lawyers consider these modules
> > > undistributable from a legal pov... I assume you have special ones
> > > that somehow think it's ok ;-)
> >=20
> > On Sat, Feb 18, 2006 at 01:25:31PM +0100, Arjan van de Ven wrote:
> > > > IANAL, and you made me worry, so I checked.
> > > >=20
> > > > There is a clear redistribution license of the binary HAL (the rest=
 is
> > > > GPL/BSD) that allows unmodified redistribution when some disclaimers
> > > > are propagated into the derived product. I see that I need to make
> > > > those disclaimers more visible in the packaging, but other than tha=
t I
> > > > don't see any legal obstructions. Do you have any details on why Red
> > > > Hat's laywers claim that madwifi is undistributable?
> > >=20
> > > I don't work for Red Hat so I can't claim anything RH's lawyers do or
> > > did say. Nor can I claim or say anything the laywers of my current
> > > employer say or did say.
> >=20
> > so which lawyers were you referring to in the quote above? Or was that
> > just a guess?
>=20
> I'm sorry but I can't disclose that. Yes that sounds lame but it's the
> way it is.

Yes, it does. It cannot be taken into account then.

> > > First of all: the rest CANNOT be GPL. Simple: The GPL requires that t=
he
> > > entire "work" is GPL, or none of it.

> > But you are confusing the license of the "whole product" with "the
> > rest". And madwifi has a very clear separation of *.o files
> > created out of GPL/BSD code and such of the HAL. So I think it's
> > correct to claim that the binary part of madwifi has a specific
> > license and the rest another. What the resulting total license is
> > is another story.
>=20
> not quite. Read the GPL: [...]
>=20
> the words and the intent are clear: If you make a bigger work which
> includes GPL components, the entire work needs to be GPL, including all
> of it's components.

OK. So you're talking about the whole work again.

> Seperate files doesn't even matter; there's ONE rpm on your site, which
> is the work, that is shown for example from the fact that the pieces of
> the rpm are not useful alone, but only when used together.
> "work" does not stop at the file boundary.

And now about the rpms (a whole work). The fact is that parts of
madwifi are under a proprietary license and parts of it are GPL. What
the resulting rpm is is a different matter, and breaking a whole work
into its components is not what is targetted here, the arguments are
about the project as it is offered at madwifi. Most importantly about
whether the licensing of the madwifi project allows one to package it
up and offer the packages under a distributable license. And to the
best of knowledge this is still the case.

So the claim remains: madwifi is shipping the code in part under a
proprietary license and in part dual licensed under the GPL/BSD. Which
means you can use the code that is dual licensed under the GPL if you
like. For a derived work like the rpm maybe the GPL is void and the
BSD license comes to a rescue. The rpm remains redistributable (under
certain terms like unmodified HAL, the no warranty disclaimers etc.)

> > > Third: The compiled module contains code from the GPL kernel, and is
> > > otherwise a derived work of the kernel, since large portions of it we=
re
> > > created with the linux kernel in mind and are only usable when used
> > > linked to the exact linux kernel they were compiled with, and which c=
ode
> > > they included in the binary. (Think of it as writing an extra chapter=
 to
> > > a Harry Potter book, since it reuses the same characters and story
> > > lines, it's clearly a derived work. Such kernel modules are very simi=
lar
> > > to such an extra chapter). This is clause 3 of the GPL.=20
> >=20
> > That would render all IHV/ISV driver contributions to RHEL4
> > illegal.
>=20
> nope.. only the non-GPL ones which are a derived work.

Let me rephrase: All *closed source* IHV/ISV driver contributions to
RHEL4 would then be illegal. So why is there a mechanism in the kernel
at all to allow for loading pure binary modules and tainting if this
is always illegal? Isn't that a contradiction? In your argumentation
"tainted" would be a synonym for illegal, and whoever allowed loading
of tainted modules in the kernel would then be aiding illegal
activities to be performed. Legally it wouldn't make sense to allow
tained modules to load at all.

> >  I.e. every Proliant with HP's Linux health monitoring, nvidia
> > graphics drivers, AVM ISDN/DSL, softmodem drivers, and so on. And not
> > only for a distributor like Red Hat, but for the IHV/ISV
> > themselves. Is that what you are saying?
>=20
> mostly. There are some exceptions which are "gray area", but by and
> large...=20

So, we can go and sue HP, SuSE, AVM and all of them, as well as the
criminal action endorsing kernel devels that allow loading tainted
kernel modules at the first place?

Anyway this drifts into the endless debates about GPL. I'll wait until
I'll get sued, too. Probably HP and SuSE will get sued before me, as
there isn't much you can get out of me ;)
--=20
Axel.Thimm at ATrpms.net