Re: Reading NI TDMS files

[John W. Eaton]

On 19-Dec-2007, Sergei Steshenko wrote:

| --- "John W. Eaton" <address@hidden> wrote:
| 
| > On 19-Dec-2007, Sergei Steshenko wrote:
| > 
| > | What you propose is exactly the way proprietary ATI/NVidia drivers
| > | are distributed.
| > 
| > I think that is "allowed" in part because the people who hold the
| > copyright on the Linux kernel do not object to it, not because the GPL
| > itself is clear on the issue, or that the FSF agree with the position
| > taken by the copyright holders of the Linux kernel.
| > 
| > | Basically, you can do whatever you want with GPL code as long as you
| > | do _not_ (re)distribute the resulting binary.
| > 
| > The FSF disagree, or at least have in the past.  As I recall, the
| > argument went something like this: suppose you would like to
| > distribute a proprietary addition to a program covered by the GPL, but
| > instead of distributing them linked together, you distribute the parts
| > separately, along with directions for how to link them together.  The
| > end result is the same, and the FSF claimed that this was infringing.
| > 
| > Again, I refer you to the GPL and the GPL FAQ, and, if your questions
| > are still not answered, please ask for clarification from the GNU
| > project and ask them to update the FAQ.

| Please pay attention to
| 
| "
| As a consequence, a company or other organization can develop a modified 
version and install that
| version through its own facilities, without giving the staff permission to 
release that modified
| version to outsiders.
| ".

Given the statement about proprietary drivers for the Linux kernel, I
thought you were referring to distributing GPL and GPL-incompatible
code separately, and having the user link them together.  That is what
I thought you were getting at when you wrote

  Basically, you can do whatever you want with GPL code as long as you
  do _not_ (re)distribute the resulting binary.

I.e., I thought you were trying to say it is OK to distribute
interface code that ultimately links Octave to some code that has
GPL-incompatible licenses, as long as the two parts are not
distributed together as a single binary.  My interpretation is that it
is not OK, because the end result is the same, and such a distribution
would just be using a convoluted method to achieve the same end result.

| And that is why I was unpleasantly surprised to see that 'octave'
| developers stopped supporting building (_not_ distributing) 'octave'
| with METIS library - after an octave version upgrade mu build
| failed, but thanks to David, who explained me publicly that adding
| "-lmetis" to LDFLAGS on 'configure' command line would solve my
| problem, I no more have the problem.

I decided to make this change because with the way things were before,
it was too easy for packagers to build binaries linked with METIS
without realizing that there was a licensing problem.

| In the end, Linux kernel is just server for whatever programs the OS runs ...

That is separate from the issue of GPL-incompatible drivers that are
linked with the kernel.

jwe