Re: [OctDev] GPL violations

[Judd S. <js...@gm...>]

On Fri, Aug 27, 2010 at 1:56 AM, David Bateman <dba...@db...>wrote:

> To me it makes sense to have to option of writing non GPLed code for Octave
> for those that have restrictions from their employers that constrain them
> from from releasing certain code to the general public and then allows these
> same employees to work with Octave and contribute back to it with other code
> that the boss doesn't consider in this same light. So Octave can win from
> this situation, but yes I agree that its shame that Mathworks also profits
> from this.
>

There are often misconceptions about what the GPL requires. The GPL does not
require that you distribute modifications or that you are even able to
distribute modifications. But, if you *do* distribute modified versions the
modifications must follow the distribution terms of the GPL.

Use *within* a company is not distribution. Even oct-interfaces can be
written for use within a company, but the cannot be distributed to
non-employees. Someone working at a company can develop code that integrates
with anything (assuming the license of the other code allows it), but they
can only distribute code to non-employees/non-contractors under terms of the
GPL. This may imply that the code cannot be distributed. Modifications (bug
fixes etc) that do not require GPL-incompatible software can be contributed
back to octave.

I think it is incorrect to say that use of MEX allows *anything* to be used
with Octave. The purpose of the MEX interface is interoperability with
Matlab. It seems requisite that any distributed MEX software must work in
Matlab and you must have access to a Matlab license to guarantee that. It
is not clear to me that Octave's MEX interface can be used to *develop*
extensions with the intention of using them in octave.

Please be aware in the change in the ToS of mathcentral a while back, matlab
> attempts to make the following changes
>
> 1) Forces a BSD license on the code there, and
>

Well, the GPL can't exactly apply in the case of GPC, can it? In any case,
the 3-clause BSD license used on MatlabCentral is GPL compatible, and the
GPL license cannot apply to code based on GPC.

where it is stated that we have no idea if there is a legal basis for such a
> change, but that we shouldn't assume that it does. If you want to use this
> code the contact the author directly rather than downloading it from
> mathcentral.


Contacting the author was offered as a courtesy (with the purpose of raising
awareness of octave), but is not required by the BSD.


> 2) Makes it impossible to use code downloaded from there with products
> other than matlab.
>

That is not true. The license on the software hosted at MatlabCentral is
three-clause BSD. By distributing their work under this license, the
copyright holder does not restrict its use to users of Matlab. In general,
Mathworks does not own the copyright to contributed software hosted on
MatlabCentral has no authority to change license terms beyond those
specified by the author (the exceptions would be files contributed by
employees). In any case, everything up there is supposed to be 3-clause BSD
licensed (I found a counterexample, where the web interface claimed the BSD
license applied but license inside the files had a non-commercial
restriction. I was not clear if the files had just not been updated to
reflect the license change).

It's unfortunate that the author of the GPC library doesn't want to work
with us, but I'm afraid the die are cast. The issue of whether the
particular files hosted on source-forge violate the GPL has been
resolved--they do and cannot be distributed. Even more troubling is that the
author of GPC advertises the availability of an octave extension on the GPC
main page.

 The copyright is explicitly in the name of H.J. Woltring with an address of
> "Philips medical systemss, Eindhoven", doesn't this make the Copyright be in
> the name of the author rather than Philips?


No, not automatically it doesn't. Copyright automatically goes to the author
at the time of creation, but employment contracts include clauses that
automatically transfer ownership to the employer.

As another example of the author not being the same as the copyright holder,
the FSF has a policy of requiring copyright transfer for all contributions.
Among other things, this allows the FSF to update its licenses without
seeking individual permission from contributors (who may be dead or
unreachable) and it gives the FSF legal standing to pursue violations in
court. Nevertheless, individual files still carry the names of the authors.


--judd