[MINC-development] experiments with EBTKS and N3

[Jordi Gutiérrez Hermoso]

On 20 August 2012 13:11, Alex Zijdenbos <zijdenbos at gmail.com> wrote:
> Well, ideological arguments aside,
[snip]
> Migrating it to a restrictive license

You did not push ideology aside as you said you would; you used
ideologically-loaded words such as "restrictive" for the GPL. :-(

> like GPL will have very practical, negative effects on part of its
> current user base (including me, but I am sure there are quite a few
> others). I don't think that should be taken lightly.

In that case, FFTPACK or maintaining the current FFT implementation
are the only reasonable choices. There is no point in writing wrappers
for GPL'ed libraries like FFTW. There is an idea that only the method
in which things communicate matters, that if you make a wrapper that
uses pipes instead of dynamic linking, then you're exempt from the
GPL. According to the FSF, this is not the case. The nature of the
data that is passed around also matters, not the method alone:

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

So, if MINC doesn't need very good FFT computations, you can use an
inferior method. It's probably easiest to follow Vladimir's original
suggestion and just maintain what is currently there.

I'm less capable of doing this myself, however. It's very easy for me
to hook in FFTW into MINC. It's less easy for me to figure out what is
wrong with the current FFT implementation.

- Jordi G. H.