Re: Distribution of media content together with GPLv2 code in one package?
On Sun, Apr 18, 2010 at 11:59:39AM +0200, Francesco Poli wrote:
> In summary, I don't agree that the author's thought is "transparently
> written down in the source code" of a program.
> At the very least, it's *not necessarily* written down.
Okay. It SHOULD be ;)
But sometimes, the "thought" involved may just be "I found this nice algorithm
in the Knuth". In case of program code, however, nobody would require the TeX
source of The Art Of Computer Programming as source code if an algorithm from
there was used. For music, a similar thing is commonly required, which
personally strikes me as odd and is why I was bringing up this comparison
between synthesis and a physical recording.
> > Something that is simply impossible for many kinds
> > of music or other media.
>
> Music may be as obscure as some computer programs.
> Music may also be as clearly expressed as some other computer programs.
>
> I think you're overgeneralizing.
Not the type of music/sound I am talking about, which is based on physical
recordings. There is no way to clearly express the sound of a real violin.
> > If you want to define source code as "whatever is needed for replicating the
> > output, in a form that can be edited", then spoken voice has no such thing.
>
> I have never defined source code like that.
> Source code is the preferred form for making modifications to the work.
>
> Make no mistake, it's *not* the preferred form for *recreating* the
> work from scratch: it's the preferred form for *making modifications*
> to it.
What if the only "common" modification of the work would be rewording? All else
you can do with a recording, is putting it through audio filters. But that kind
of editing is HIGHLY limited and not powerful enough for common needs.
> If you want to *modify* the recording of spoken voice, some form of
> recording will surely be preferred over other formats.
> That is the source.
In that case, this would be the finished audio file, as no better form exists.
One might argue about lossy or lossless encoding of the data, but personally
I'd even accept a high-quality ogg file as viable form for editing, and
sometimes when I make my own recordings, I only save that (but then at a high
level, 7 or above). In case I want to do further edits BASED on that recording,
I can just as well use the ogg file.
> I cannot be more specific, without discussing the details of a given
> practical case; and I am *not* willing to dive into the details of a
> practical example, since that would drive the discussion far away...
Assume the recording is spoken words for a tutorial level of a game.
The typical "editing" performed to that is changing the wording. And there is
no form of it that CAN be used to edit it - if someone who is not able to sound
like the original speaker is going to e.g. add a part to that tutorial level,
he will have to rerecord ALL sounds from scratch.
Assume we will call THAT "editing", then the "text" spoken in the recordings is
as close to source code as we can get. However, the GPL explicitly requires
machine-readable source code, and an ASCII text file containing the lines to be
spoken certainly isn't meant by that (as otherwise, a low quality scanned-in
jpeg of a printout of the source code of a program would be just as
acceptable).
I'd go even further now, and ask: CAN such a work even be put under the GPL?
Does maybe the type of work explicitly exclude licensing that requires
machine-readable source code, as no such thing exists? Doesn't that mean that
a game containing such a work CANNOT be put under the GPL as a whole, and that
it HAS to be separated out into multiple packages and be made in a way that no
code or other asset covered by the GPL explicitly refers to the voice
recordings, as that may be considered linking?
Doesn't this mean that the ONLY sound files containing spoken voice that are
eligible for Debian main are sound samples synthesized using a free speech
synthesis engine (like espeak), and that both recorded voice (which is highly
opaque and uneditable) and commercial voice synthesis are ruled out by Debian
main?
I simply can't seem to find a reason why an opaque recorded voice sample should
be considered "more free" than the output of a commercial voice synthesis
engine. Any edit you can do on the recorded voice sample, you also can do on
the synthesized output using the very same tools. The only difference is that
with the synthesized audio sample, you have an additional way to edit the file,
which requires that commercial application.
On the other hand - IF we accept the synthesized audio file as DFSG-compliant
and GPL-compliant, what speaks against use of commercial samples in music?
Musicians tend to understand musical score notation as the "source code" of
music, and that can still be easily provided (e.g. in form of a project file
for some application, or a MIDI file). It just cannot be used to generate
identical output without access to the commercial samples (just like the C
source code cannot be used to generate identical output without access to the
compiler, which may be commercial too - but that GPL and a commercial compiler
don't conflict has been accepted for long). Nothing prevents you from
performing it in a real orchestra, or using your own samples in place of the
"missing" ones - just like nothing prevents you from reading out the same lines
of text with YOUR voice, and recording that.
Rudolf Polzer
Reply to: