Re: GPL - GSL and derivative work.

Re: GPL - GSL and derivative work.

[Timothy H. Keitt]

These are pretty clearly spelled out at www.gnu.org.  Short answer: 
LGPL, linking from closed-source OK.; GPL, you have to release the 
source of the application under GPL to link with GPL library.  In any 
case, the copyright holders are free to issue you an alternative license 
if they so choose.

T.

Nicolai Hanssing wrote:

>Dear listmembers
>
>I was wondering what the authors of gsl, consideres to be derivative work.
>An example follows:
>
>--------
>A firm writes an application, that uses/links with gsl. The firm does not
>however modify the gsl-source.
>
>Now the firm wants to sell/distribute the application, then what is forced
>by the GPL?
>
>The firm,
>*) Has to make the source of the application publicly available?
>I.e. will rewrite the application to not use GSL .... :-)
>*) Has to distribute the source of gsl along with the application?
>*) Has to include the GPL-license in the documaentation?
>*) Must not advertize with the use of GSL without permission?
>*) anything else...
>
>I'm currently writing my thesis in cooperation with a firm, and I'm using
>gsl, so what are the firm's possibilities afterwards [of course the source
>is public - it's a thesis] for future use?
>
>
>Regards
>	Nicolai Hanssing
>	Denmark
>

-- 
Timothy H. Keitt
Department of Ecology and Evolution
State University of New York at Stony Brook
Stony Brook, New York 11794 USA
Phone: 631-632-1101, FAX: 631-632-7626
http://life.bio.sunysb.edu/ee/keitt/

Nelder-Mead Simplex

[Nicolai Hanssing]

Are anyone working on other multidim minimisation-routines?

Otherwise, you can expect me to atleast implement SIMPLEX in
gsl_multimin_, i.e. a simple slow non-gradient solver.  Comments are
welcome.

Regards
	Nicolai Hanssing
	Denmark

GPL - GSL and derivative work.

[Nicolai Hanssing]

Dear listmembers

I was wondering what the authors of gsl, consideres to be derivative work.
An example follows:

--------
A firm writes an application, that uses/links with gsl. The firm does not
however modify the gsl-source.

Now the firm wants to sell/distribute the application, then what is forced
by the GPL?

The firm,
*) Has to make the source of the application publicly available?
I.e. will rewrite the application to not use GSL .... :-)
*) Has to distribute the source of gsl along with the application?
*) Has to include the GPL-license in the documaentation?
*) Must not advertize with the use of GSL without permission?
*) anything else...

I'm currently writing my thesis in cooperation with a firm, and I'm using
gsl, so what are the firm's possibilities afterwards [of course the source
is public - it's a thesis] for future use?


Regards
	Nicolai Hanssing
	Denmark

Re: Nelder-Mead Simplex

[Brian Gough]

Nicolai Hanssing writes:
 > Are anyone working on other multidim minimisation-routines?

No, I don't think so.

 > Otherwise, you can expect me to atleast implement SIMPLEX in
 > gsl_multimin_, i.e. a simple slow non-gradient solver.  Comments are
 > welcome.

Cool.  

Make sure you can release it under the GPL (for example, don't copy
from Numerical Recipes).

The GSL 'coding standards' are in doc/gsl-design.texi.

Re: GPL - GSL and derivative work.

[Edwin Robert Tisdale]

Mark Galassi wrote:

> There is a discussion of these issues on the GNU web pages
> which matches what I say
> and is a legal opinion backed by the FSF's legal counsel.
> 
> If another identical API exists then you have a grey area.
> People have done the trick (in other software)
> of writing a low-quality compatible API
> that was public domain instead of GPL'd

Yes.  The worst tricksters are the FSF and the GNU project.
The GSL is, at best, simply another implementation of
the Basic Linear Algebra Subprogram (BLAS) library,
The Linear Algebra PACKage (LAPACK) and Numerical Recipes in C.

You can't force anyone to distribute the source code
for their application programs
simply because it uses the same API as the GSL.
If you could, then proprietary library vendors could require
a royalty from every program that used their API
even if it was linked to a GPL'd implementation.

> Regarding your comments on the use of the GPL instead of the LGPL,
> we are a GNU project and there are very clear reasons
> for the GNU project to promote using the GPL
> instead of the Lesser GPL for almost all libraries.
> See:
> 
> 	http://www.gnu.org/licenses/why-not-lgpl.html

There are better reasons to promote the LGPL instead of the GPL
for numerical libraries like the GSL.
But applying the GPL instead of the LGPL
is just one of the reasons why the GSL is doomed.
The GPL unnecessarily encumbers the GSL.
It does not make the GSL or applications that use it more free.

I believe that you mean well Mark but your attitude
is one of the worst enemies of free software.

Re: GPL - GSL and derivative work.

[Edwin Robert Tisdale]

Mark Galassi wrote:

> On this list, I simply want to assure anyone that
> if your software can only work with GSL
> then your software is subject to the relevant parts of the GPL.

Nonsense!

The GPL applies only if you distribute your software
any only if you distribute the GPL'd library (GSL) with it.
You cannot copyright the GSL API.
Anybody can implement the GSL API.
The Free Software Foundation and the GNU project
exist, in part, because they have succeeded
in re-implementing existing APIs.
If it was possible to copyright an API,
there would be no FSF or GNU project.
The FSF would only undermine it's own right
to implement and distribute software using existing APIs
if it were to take other programmers to court
simply for using the GSL API.

Applying the GPL to a library like the GSL
is simply an attempt to stifle competition
in just the way that Microsoft attempts
to stifle competition.

Re: GPL - GSL and derivative work.

[Edwin Robert Tisdale]

Timothy H. Keitt wrote:

> These are pretty clearly spelled out at www.gnu.org.
> Short answer: LGPL, linking from closed-source OK.;
> GPL, you have to release the source of the application
> under the GPL to link with a GPL'd library.

The GPL only prevents you from distributing the GPL'd library
without distributing the application source code.
You can distribute the application object code
without the application source code without violating the GPL
and let users link it into the GPL'd library themselves.
I don't think that there is any practical way for the FSF
or the library developers to prevent users from doing this.

Re: GPL - GSL and derivative work

[Edwin Robert Tisdale]

Mark Galassi wrote:

> Once you reimplement GSL, you can say that
> the software does not depend on GSL for its functioning.
> 
> But until you do, if you distribute software and say,
> "This will only work with GSL
>  because no other implementation of that API exists."
> then you have a derivative work.

Suppose that the court agrees with you, counselor,
then someone reimplements the GSL API.
Would we still have a derivative work?
Would other applications using the GSL API
but written after the GSL API was reimplemented
then also be derivative work?

Re: GPL - GSL and derivative work.

[Mark Galassi]

    >> On this list, I simply want to assure anyone that if your
    >> software can only work with GSL then your software is subject
    >> to the relevant parts of the GPL.

    Edwin> The GPL applies only if you distribute your software any
    Edwin> only if you distribute the GPL'd library (GSL) with it.

Of course: I said that you are subject to the relevant parts of the
GPL, so it only applies if you distribute your software.

    Edwin> You cannot copyright the GSL API.  Anybody can implement
    Edwin> the GSL API.

Of course, and once you reimplement GSL then you can say that the
software does not depend on GSL for its functioning.

But until you do, if you distribute software and say "this will only
work with GSL because no other implementation of that API exists" then
you have a derivative work.

Re: GPL - GSL and derivative work.

[Steve ROBBINS]

On Tue, Aug 14, 2001 at 07:55:50AM -0700, Edwin Robert Tisdale wrote:
> Timothy H. Keitt wrote:
> 
> > These are pretty clearly spelled out at www.gnu.org.
> > Short answer: LGPL, linking from closed-source OK.;
> > GPL, you have to release the source of the application
> > under the GPL to link with a GPL'd library.
> 
> The GPL only prevents you from distributing the GPL'd library
> without distributing the application source code.
> You can distribute the application object code
> without the application source code without violating the GPL
> and let users link it into the GPL'd library themselves.

This is not true.  If your application requires the GPL'd library
it is to be considered a "derived work" of the library, and all
sources must be shipped with a binary.  This is pretty clearly
laid out in the text of the GPL, and the GNU web site has further
disucssions.  In particular, see

  http://www.gnu.org/licenses/gpl-faq.html#IfLibraryIsGPL


If you intend to split legal hairs about what constitutes
a "derived work", that is a discussion I do not want to have.
Try the newsgroup gnu.misc.discuss.


> I don't think that there is any practical way for the FSF
> or the library developers to prevent users from doing this.

Well, the practicality is a separate question.  ;-)
But it is beyond question that this is not what authors of
GPL'd libraries *intend*.

-Steve

Re: GPL - GSL and derivative work.

[Mark Galassi]

    Edwin> Yes.  The worst tricksters are the FSF and the GNU project.

This is a GNU project mailing list, so such comments are quite
inappropriate.  Please stop making them.  There are other places for
it.

On this list I simply want to assure anyone that if your software can
only work with GSL then your software is subject to the relevant parts
of the GPL.

Re: GPL - GSL and derivative work.

[Edwin Robert Tisdale]

Mark Galassi wrote:

> E. Robert Tisdale wrote:
> 
> > You can distribute the application object code
> > without the application source code without violating the GPL
> > and let users link it into the GPL'd library themselves.
> 
> Someone else has pointed it out but let me repeat that
> the above paragraph is completely wrong.
> 
> If the GNU Scientific Library is *required*
> for your program to function,
> then your program is a "derivative work"
> and must be distributed as spelled out by the GPL
> (the easiest way is to put your program under the GPL)
> even if you don't ship a binary that is statically linked
> to the GNU Scientific Library.
> 
> If the GNU Scientific Library is just one of many libraries
> with identical APIs, then you might have other options.
> 
> > I don't think that there is any practical way for the FSF
> > or the library developers to prevent users from doing this.
> 
> Breaking the law is breaking the law
> even if it is difficult to get caught.
> Breaking the law to exploit people's volunteer work
> and not return to the free software community
> counts as unethical in my book.
> 
> All things considered, trying to "work around" the GPL is dumb.
> People could try a copernican-style shift in their thinking:
> the GPL is a powerful tool
> and people should look at the "derived work" clause
> as an opportunity to try a licensing model
> that they would have ignored otherwise and
> that might actually work much better for them.

Your personal sense of "justice"
is not an adequate substitute for law.

An application program does not "require" the GSL
just because it uses the same API.
I don't think that you can copyright an API.
Anyone could implement the GSL API
without infringing on the GSL copyright.

The GSL developers can't make law
by simply including the a GPL notice
or expressing the authors' "intent".
Law is made when the license is challenged in court
and the court decides how the copyright law
and the licensing agreement is to be interpreted.
I don't believe that has happened yet in this case.

In my opinion, the GSL developers made a poor choice
when they decided to use the GPL instead of the LGPL
and I would have no sympathy for them
if some application developer decided to work around it.
On the other hand, I believe that the GSL is a poor choice
for any application development let alone commercial applications
and I would have no sympathy for any application developer
who used the GSL.

Re: GPL - GSL and derivative work.

[Mark Galassi]

    Edwin> An application program does not "require" the GSL just
    Edwin> because it uses the same API.  I don't think that you can
    Edwin> copyright an API.  Anyone could implement the GSL API
    Edwin> without infringing on the GSL copyright.

There is a discussion of these issues on the GNU web pages which
matches what I say and is a legal opinion backed by the FSF's legal
counsel.

If another identical API exists then you have a grey area.  People
have done the trick (in other software) of writing a low-quality
compatible API that was public domain instead of GPLed

Regarding your comments on the use of the GPL instead of the LGPL, we
are a GNU project and there are very clear reasons for the GNU project
to promote using the GPL instead of the Lesser GPL for almost all
libraries.  See:

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

Re: GPL - GSL and derivative work.

[Mark Galassi]

    Edwin> You can distribute the application object code without the
    Edwin> application source code without violating the GPL and let
    Edwin> users link it into the GPL'd library themselves.

Someone else has pointed it out, but let me repeat that the above
paragraph is completely wrong.

If the GNU Scientific Library is *required* for your program to
function, then your program is a "derivative work" and must be
distributed as spelled out by the GPL (the easiest way is to put your
program under the GPL), even if you don't ship a binary that is
statically linked to the GNU Scientific Library.

If the GNU Scientific Library is just one of many libraries with
identical APIs, then you might have other options.

    Edwin> I don't think that there is any practical way for the FSF
    Edwin> or the library developers to prevent users from doing this.

Breaking the law is breaking the law, even if it is difficult to get
caught.  Breaking the law to exploit people's volunteer work and not
return to the free software community counts as unethical in my book.

All things considered, trying to "work around" the GPL is dumb.
People could try a copernican-style shift in their thinking: the GPL
is a powerful tool, and people should look at the "derived work"
clause as an opportunity to try a licensing model that they would have
ignored otherwise and that might actually work much better for them.