[darcs-users] Re: how to redistribute darcs+Eclipse

[Jamie Webb]

On Thu, Jun 09, 2005 at 09:11:43AM +0200, Ketil Malde wrote:
> Jamie Webb <j at jmawebb.cjb.net> writes:
> 
> >>>> A patch is considered a derived work,
> 
> >>> Are you sure?
> 
> > Not absolutely. There appear to be no precedents specifically relating
> > to diffs. The similar cases that get cited found that third party Duke
> > Nukem levels are considered derived works, and likewise fan fiction
> > relating to films and books, even though no part at all of the
> > original was copied. The argument is that the new works would not have
> > been possible without the originals, and hence that the 'spirit' of
> > the original was in some sense derived from.
> 
> Impressive.  I've always thought that copyright covered the literal
> work, and not its "spirit" or ideas or such.

No. For example, translations of texts and cover versions of music are
definitely derived works. Ideas in the technical sense are not
copyrightable (they are patentable), but more artistic ideas (e.g. a
character in a story) are.

> An interesting
> consequence of this is that parodies are only possible if the parodied
> part agrees to it.

Copyright law has specific exceptions for derived works created for
the purpose of journalism or parody. It's considered 'fair use'.

> Wow.  How did the creative communiteis paint
> themselves into such a legal corner?

Probably, they didn't. The evil corporations which feed off the
creative community did it for us.

> And - doesn't this mean that any program developed with Visual Studio
> and requiring Windows to run, is partial copyright Microsoft?  It
> seems fairly analogous to me.

I don't think that's the same thing. The program happens to run on
Windows, but from a creative perspective essentially the same program
could have been developed with any tool targeting any OS.

Clearly there's a line to be drawn, but I think we can be fairly
certain that an OS and its applications are distinct works. I don't
believe there's any case law that can help here though.

[snip SCO stuff because I didn't follow it]

> > As for a suit, even were patches not considered derived works, the law
> > set great store by 'accepted practice' and the actions of the
> > 'reasonable man'. Posting a patch against a GPLed program to a public
> > list dedicated to discussing that program would cause most in the
> > industry to infer that the patch was GPLed, so we can expect a judge
> > to agree. But, it could be expensive.
> 
> Of course most would infer that.  A lot of people infer that stuff on
> they find on the internet is theirs to use.  I'm not sure public
> perception defines reality, though. :-)

True. We find ourselves at the mercy of common sense.
 
> I would hope that users who incorporate such a patch would be in "good
> faith", and not liable for damages, but I also think that if there is
> no explicit licensing, the copyright holder could require the patch be
> removed.

Unfortunately, regardless of the legal position, you're probably right
from a financial one.

> (I may be totally off base here, but if not, this could be
> an argument for the "pull" model rather than "send" -- "pull" would of
> course get its changes from a repository that includes the licencing
> information, the "send" is just the patch.) 

I don't think there can be a difference. Two possibilities:

- A repo is legally a single work and all patches are derived works of
  their context. In that case, just sending would be sending a derived
  work.

- Each patch is an independent work of its author, and a repo is just
  a vessel to hold them all (it happens to hold them in combined
  format in working and pristine, but that's an implementation
  detail). In that case, neither pull nor send would be sufficient,
  since either way the patch itself contains no license info and none
  is transmitted.

Maybe Darcs could make things watertight with a 'license' pref that
gets verified as part of apply?

I.e. if the license pref is set, prior to recording a patch you get a
message:

The recorded license for this patch will be: GPLv2+. See COPYING.

And during apply, there would be a warning if the license texts do not
match (but you could proceed anyway, e.g. to incorporate LGPLed code
into GPLed).

-- Jamie Webb