Scope of copyright on derivative works

Philippe Verdy verdy_p at wanadoo.fr
Sat Sep 29 15:05:41 UTC 2007


Alexander Terekhov [mailto:alexander.terekhov at gmail.com]wrote:
> On 9/29/07, muddying the water, Lawrence Rosen <lrosen at rosenlaw.com>
> wrote:
> > Alexander Terekhov erroneously suggested:
> > *    Copyright (c) 2007 GPL Developer Who Made Changes <gpl at example.org>
> > *
> > *    This file incorporates modifications covered by the GPL... [blah
> blah]
> >
> > Your "blah blah" is wrong. The correct notices would be:
> >
> >   Copyright (C) 2007 GPL Developer Who Made Changes.
> >   This file is licensed under the GPLv2 [or GPLv3].
> >
> > I'm serious: This [entire!] file is licensed under the GPLv2 [or GPLv3]!
> 
> I'm also serious: only modifications are licensed under the GPL, not
> the entire file.

Correct, but only if all the rest was not licenced under GPLv2 [or GPLv3],
for example if the rest was licenced under BSD licence, or is now Public
Domain.

However, the resulting combined work is covered by two licences, and as long
asit remains under this form, the licences are not separable. If the changes
are not identifiable and not easily separable, then the two licences MUST be
compatible with each other; otherwise the possible contradictions between
the two licences would be unsolvable.

That's why we need rules to determine which licenses are *compatible* with
each other, notably because most software are made now as combined works
from multiple origins, and they constantly evolve in each of their parts;
upto the point that the various pieces will become intimately linked to each
other and non separable. But even in this case, the original licenses must
still be preserved (or possibly upgraded into a newer one if this right was
granted by the component licences, or if the original authors authorizes it,
another reason why we must keep all copyright assignments to each part).

> The act of sublicensing is what happens when a licensee becomes a
> licensor to some other party by granting some or all of the rights
> that they received as a licensee.  The recipient of BSDL'd material
> gets a license from the original licensor -- without the middleman
> getting a chance to do anything at all regarding granting some or all
> of the rights that middleman received as a licensee. The middleman may
> or may not grant rights to his modifications though.

Correct. The same is true with the GPL which allows no middleman
interference between a licensee and the original authors. The only allowed
interferences are those related to the copyright assigned to the author of a
substantial modification whose content is copyrightable (but then the
original authors will request that these changes become identifiable in a
way that the original parts that are kept remain assigned to the original
copyright holder).

If you create a simple combined work without identifying the relevant parts
whose parts are clearly identifiable to a copyrightable author, this change
may not be authorized, even if you list separately the complete list of
copyright assignments, because all parts (whatever their source) would be
covered by multiple licences. This will be clearly a problem if those
licences contain contradicting terms between each other. That's another way
to explain that it's best not to intimately mix several parts with different
licences, if they are not compatible with each other.

The proliferation of licences in the open-source or free software movements
is becoming a problem. At least we have two (separate) efforts trying to
determine which licenses are compatible with each other (OSI and FSF). But
this is still not enough. The original licence authors should sign an
agreement permitting such compatibility claims.

Another possibility is to make a joint effort to request that authors
upgrade their existing multiple licences into a common licence with similar
rules. That's what the FSF has done since long, but the OSI does not seem
ready to accept this, and still militates for a proliferation of
"open-source" licences which may be compatible with a single "reference"
licence (the OSI conditions), but not compatible with each other in any
other combinations. Nowif you try mixing parts covered by 3 or more
OSI-approved licences, we are in serious troubles when trying to determine
what to to.

What this means: there will be only two classes of licences tolerated: the
GPL-derived licences, and the BSD-like permissive licences.

Plus the "Public Domain" which is not a licence but is defined externally by
default, and by various national laws, meaning that all references to
"Public Domain" only are invalid if they are not associated to a national
law defining it precisely, given that there's no copyright notice required
or preserved in the "Public Domain". Such references to "international
public domain" are invalid (unless they are defined later by international
WIPO treaties, and given a distinctive name according to these international
treaties):

This was demonstrated by some recent changes in the US and Russian
legislation, extending the initial duration of exclusive authors rights even
AFTER these rights had already been terminated years before, forcing some
contents that had fallen to the public domain to come back to their past
status of copyrightable and licenciable contents! Consider the "Public
Domain" as a licence granted temporarily by the governement instead of by
authors.

For these reasons, in my own projects, I now reject all public domain
contributions unless they are contents widely available since centuries (at
least 100 years old) from many sources, and I consider those recent
creations in software tentatively published in the "public domain" as being
published with a completely invalid licence, most probably illegal.






More information about the License-discuss mailing list