[darcs-users] Licensing and copyright fun.

[Stephen J. Turnbull]

Trent W. Buck writes:
 > trentbuck at gmail.com (Trent W. Buck) writes:
 > 
 > > If a file has no license declaration, it *is not licensed*.
 > > Therefore adding license declarations is REQUIRED.  It is *not*
 > > sufficient to simply include a COPYING file in the root
 > > directory.

This isn't true, in principle.  In the U.S., the licensor doesn't even
need to provide a COPYING file, or even say a word.  Violating
copyright is not a crime (in most cases, and even when it is it is
unlikely in the extreme to be prosecuted until a copyright holder
complains).  All they need to do is *not sue you*.  In the case of
copyleft open source, you also have good reason to suppose that there
is an implicit license (see below).

In the practice of open source, however, this requires extreme trust
on the part of licensees, especially since (as LZW amply demonstrates)
such "implicit licenses" can be revoked at any time, leaving the users
high and dry with respect to legal extortion.

 > > This applies to *ALL* files (that aren't autogenerated),

I don't see any reason in copyright law for autogenerated files to be
excepted.  For example, under a BSD license you certainly could omit
configure.in and distribute configure only, in which case you'd want
configure to contain an appropriate notice.  I think under the GPL you
would be on shaky ground to do that because of the requirement to
distribute "build scripts" as part of the whole work and source for
everything distributed in binary, of course, but you never know what
people will do with code.

 > > It's not clear to me if we can add a license declaration to a
 > > file without checking with the copyright holder.

You can, but this is where "risk to licensees of depending on implicit
licenses" enters.

 > > Are contributions to darcs

 > > - implicitly "GPL2";
 > > - implicitly "GPL2 or higher"; or
 > > - not implicitly licensed?

I would say they are implicitly GPL2, for sure, with "or higher" if
and only if that's David's license.  This is (a) the usual community
practice, and (b) implied by them publicly posting (ie, distributing)
their patches based on copyleft code to a copyleft project without
disclaiming a license.  However, that can be explicitly disclaimed at
any time by the copyright holders, as with the OpenSSL exception that
some contributers have so far not signed on to.  The OpenSSL license
is in explicit contradiction to the GPL, so I personally would
consider it impolite to depend on an implicit license with respect to
it, but I think the implicit license is still there except for
explicit refuseniks.

Unless there is a deliberate saboteur in the bunch, the principal risk
is that somebody will refuse to allow their code to remain in Darcs.
A deliberate saboteur might request damages or even call for a
criminal case, but even there, it's highly unlikely that a U.S. court
would consider there to be any crime or even damages involved, so the
only thing you'd have to do is remove the code.

 > Unfortunately, release/openssl_ok indicates that correspondents
 > were agreeing to exceptions, but not an actual license!  I think
 > this means we need to talk to everyone who has worked on Darcs so
 > far, and get them to agree to license their work under the terms of
 > GPL-2 or higher, with the exceptions mentioned in
 > release/openssl_ok.

This would be a good idea, but I see no reason why it can't be done
over time.  Unless you're that afraid of copyright sharks, in which
case my advice is "swim away!" because blood's already in the water.

There are a couple of authorities for this kind of unilateral
relicensing.  First, the Wikipedia removed some DFSG-unfree terms from
its instance of the GNU Free Documentation License (cover texts or
invariant sections or something like that).  Also, Eric Raymond wrote
a screed on unilateral relicensing, I suspect with the advice of the
OSI's lawyers: http://www.catb.org/~esr/Licensing-HOWTO.html#id2790762.