[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: GPL and linking (was: Urgently need GPL compatible libsnmp5-dev replacement :-()

On 5/10/05, Batist Paklons <batist@gmail.com> wrote:
> On 10/05/05, Raul Miller <moth.debian@gmail.com> wrote:
[a painfully inadequate redaction of my argument, with a non-rebuttal]

> I do not believe this to be a contradiction. Collective works and
> derivative works are two entirely different concepts in copyright law,
> that nonetheless can apply to the same work. In their definition their
> have nothing in regard with each other, and are both grounds to grant
> copyright protection (or authorship, as is the more continental
> approach to copyright). But that they are not related in definition
> does not bar the possibility that in fact they can be applied to the
> same work. This would lead to a double copyright protection, but if
> you look at it from the viewpoint of authorship, this simply means
> there are two different reasons why one can be considered an author of
> an original work.

Not quite.  The fact that their definitions are historically unrelated
-- one apportions the rights of original author and
translator/redactor/adapter, while the other recognizes the editorial
contribution of the publishing channel -- implies neither overlap nor
disjunction.  But the text of their Berne Convention and 17 USC (1976
and later) definitions does (to my satisfaction under US law, IANAL),
as I exhibited in the post that Raul savaged.

It makes little difference in a "normal" copyright infringement
proceeding that "Bambi Meets Godzilla plus Bilbo Potter -- Together on
One DVD!!!" is a collection of two works, each of which is a
derivative work of two works; the holders of all seven possible scopes
of copyright have separate statutory causes of action if I pirate that
hypothetical work (which would fill a much-needed gap in the
literature).  That's because I have _copied_ a work with regard to
which they all hold some statutory right to protest unauthorized

The licensing pattern by which the creation of this magnum opus can be
authorized is more complex.  It forms an inverse pyramid with seven
nodes in three layers -- of 4, 2, and 1 -- and arrows (permission
grants) pointing down to each node from _all_ its ancestors.  Those
nodes are not works; they are _copyright_holders_.  Remember, in the
US at least (where a lawyer I am not), a (non-exclusive) copyright
license (i. e., a provision in a contract -- written, oral, or implied
by conduct) is an enforceable promise not to sue under 17 USC -- no
more and no less.  (It is different in subtle ways from a "unilateral
grant of license", which is a form of estoppel, and which I'm not
going to discuss at the moment.  It has nothing to do with license
under the GPL, in which there are ample grounds to find consideration
and acceptance; that horse was beaten to a pulp on debian-legal four
or five months ago.)

Imagine ASCII art here.  No, not _that_ ASCII art!  Imagine the
directed graph described above, with seven nodes and ten arrows.  Here
are the nodes, anyway:

   Disney       Toho      Tolkien     Rowlings
         Newland               I-Hope-Not

Now licenses are terms in contracts, and under common law (AIUI,
IANAL, TINLA) contracts have two (and only two) parties.  (A document
that says it's a three-party contract is really evidence with regards
to the terms of three pairwise contractual relationships, and will be
construed accordingly if it gets to a courtroom.)  To get license from
someone, I need to bind them to a contract.  I'm going to mention
agency below, but that doesn't really move the arrows around, even if
one of the copyright holders also holds agency from another to
sublicense a relevant right.  It just means the agent has the
authority to offer to bind the copyright holder not to sue under
certain circumstances, and to accept return terms within some set of
constraints.  The agent may even be authorized to receive and pocket
the payment contracted from licensee to copyright holder.  The
associated arrow in the graph just acquires an annotation that
explains how the copyright holder came to be bound to the licensee.

(By the way, copyright assignment isn't part of this calculus; its
statutory basis is in 17 USC, totally independent of contract law.)

Now God-Save-Us needs license both to _copy_ and explicitly to
_collect_ (anthologize) two works; each license to anthologize needs
to be obtained separately from three copyright holders.  Careful --
if, for instance, agency to sub-license for purposes of anthology has
been granted to Newland by Disney and Toho (fat chance), then there
may be one document signed by Newland and God-Save-Us, but it still
establishes three contractual relationships, one in which Newland
binds himself and two in which he acts as the agent of another
copyright holder.  Newland and I-Hope-Not, in turn, each needed
license to _copy_ and explicitly to _derive_from_ (more likely to be
stated simply as "adapt" or "parody") two works.  6 + 4 = 10 arrows. 
So far so good?

Newland may have acquired license from Disney to create his specific
work and to authorize anthologies of the result.  (I'm making this up
for purposes of example, people, don't take this as the history of the
actual Bambi Meets Godzilla, about which I know nada!)  But that
doesn't imply license to anthologize the original Bambi or to create
or to authorize the creation of some other derivative work of Bambi. 
Even a blanket license to "create derivative works of Bambi" isn't
agency to authorize others to do so, unless a court determines that
there is a firm basis to construe this authorization through the
conduct of the parties.  (That basis may be as simple as "X was a
moneybags, not a screenwriter, of course both parties realized he was
going to hire somebody" -- but it's not automatic.)

(I am skipping the chain-of-derivative-works case for simplicity, but
Humberto would perhaps be right that legal authorship could
technically survive a chain of custody in the course of which all
vestiges of the first author's expression had been lost -- if it
weren't for the "de minimis" defense and the practical effect of the
abstraction-filtration-comparison test.)

Now in the Real World (TM) it's often not clear quite where to set the
bounds of implied agency when you're dealing with Hollywood, which is
a black box that eats money and spits out movie holding companies
(which are black boxes that eat gross receipts and spit out tax
losses).  The software world is worse, and the free-for-all of an open
source project without a copyright assignment policy (which is one
thing that absolutely has to be accepted in written form) is worst of
all.  But if you can tell a box of chocolates from a chocolate bar
without staying up nights worrying where the vanilla bean went, you
can distinguish between "selection and arrangement" and "translations,
adaptations, ... and other alterations".  Hypothetical persons who
want to tell me what my-pet-dictionary.com says about "arrangements of
music" need to go discuss UN-altered REPRODUCTION and DISSEMINATION
with Robert McElwaine.  (Eschew!  Eschew!)

Did I mention that "work based on the Program" = "either the Program
or any derivative work under copyright law"?  So if the Lord of the
Rings were GPLed, then release of Bilbo Potter (a derivative work)
would be permitted (if at all) only under GPL terms, while Bambi Meets
Godzilla could share a DVD with it and remain under some other license
with its own nefarious agenda.  Probably Humberto is right that the
"mere aggregation" clause is likely to be the biggest piece of
evidence that a judge uses to construe implied agency to authorize
anthologies of GPL material.  I now vacillate between implied agency
and estoppel, "mere aggregation" and "the GPL doesn't 0wn my distro /
living room", probably because IANAL and I need to be getting more

> A derivative work is not a collective work and vice-versa, but one
> work can be both at the same time.

Er, see above.  :-P

- Michael

Reply to: