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

Re: GPL as a license for documentation: What about derived works?

Disclaimer:  IANAL, IANADD, I haven't been actively engaged with
debian-legal for very long, and my interpretation of the meaning of
"derivative work" and its consequences for the scope of the GPL
appears to contrast rather strongly with the FSF's and with some other
debian-legal participants'.  But I can point you to some US case law,
and statutes elsewhere, which I believe to be relevant.  Perhaps this
will be helpful to you in assessing the quality of the local
conventional wisdom.

On Fri, 28 Jan 2005 21:49:08 +0100, Frank Küster <frank@debian.org> wrote:
> Let's assume a piece of technical documentation (standalone, i.e not
> part of a software package; something like selfhtml or LaTeX's lshort),
> is licensed under GPL, with an additional text stating what the
> preferred form for modification is (say, LaTeX or docbook).
> As I understand it, anybody can take the text and publish it as a
> printed book, as long as he also prints where the source code can be
> found (GPL clause 3.b or 3.c). If he creates a derived work - for
> example by extending each chapter, but keeping the structure and most of
> the original text - he has to license that under GPL, too (and thus
> provide the source code).

That's correct.  Publishing this book, or in general any action that
would normally require a copyright license and isn't accompanied by a
strong claim of alternate licensing terms, is fairly strong evidence
of "acceptance through conduct" of the offer of contract contained in
the GPL, in any jurisdiction where a non-exclusive copyright license
can be created by this form of acceptance.  That would include any
jurisdiction where any retail software license is enforceable.

There is a non-zero threshold for evidence of acceptance, however. 
See Specht v. Netscape, which held that a mere "browse-wrap" license
on downloadable software was similar to an open box containing copies
of a free newspaper, and that merely taking and using that newspaper
doesn't prove acceptance of the attached license.  Picking up a copy
of a free newspaper doesn't convey license to republish its contents,
however; so publishing printed copies of a GPL document is clearly
well over this threshold.

Judging from the two cases I have been able to find in which the GPL
has come up in US court (namely, Progress Software v. MySQL and
Planetary Motion v. Techplosion), there is no reason to think that
there is any defect in the GPL as an enforceable offer of contract
regarding material on which the copyright holder retains its
ownership.  Specifically, publication of a work with the GPL attached
is not release into the public domain and does consitute "use in
commerce", according to an argument in Planetary Motion (a valid
appellate precedent) which appears to me also to prove sufficient
consideration on both sides to form a bilateral contract.

Notwithstanding the public claims of the FSF's general counsel, there
is no such thing as a "non-contract license" in any jurisdiction I
have heard named.  With help from other debian-legal participants,
I've verified this assertion against statute and/or precedent in at
least the US, Canada, UK, Belgium, and Germany.  I have also asked
licensing@fsf.org about their stance, and received nothing but
pointers to an interview with the general counsel and the 1709 Statute
of Anne.  The FSF's general counsel submitted an expert witness
affidavit in the MySQL case, but it makes no reference to relevant
modern legal precedent.

In general, this means that a claim of copyright infringement cannot
succeed unless the copyright holder can demonstrate, in addition to
the usual criteria for infringement, that the defendant either
exceeded the scope of the license granted in the contract or breached
the contract to an extent that justifies rescission of the license. 
This is consistent with the judge's ruling in Progress Software vs.
MySQL, in which MySQL did not succeed in obtaining a preliminary
injunction on copyright grounds.

> I have two questions
> 1. The first is whether there are any established criteria by which the
>    creation of a derived work can be distinguished from mere aggregation.
>    I assume that if a book on the technical aspects of computer
>    typesetting would include, as an appendix, a GPL'ed text on
>    typography, this would be only aggregation. At least if typographical
>    questions don't play a role in the rest of the text.
>    But what if there are extensive references to specific parts of the
>    appendix in the text? What if it is a chapter in that book?

I'm going to try to answer the question I think you are asking: 
"Joe-Bob writes a book about GPL Work X, on which the copyright is
held by Sue.  He includes Work X as Chapter X (or Appendix X) of his
book, and makes extensive references to Work X in Chapter Y (and
possibly every other chapter).  Can Sue successfully sue Joe-Bob to
either block distribution of the book or to require release of the
'source code' for the entire book under the GPL?"  I believe the
answer to be No.

For reasons that I discuss below, I don't think that the language of
the GPL, taken together with applicable law and precedent, should be
read to cover "collections" or "compilations", irrespective of how the
undefined term "mere aggregation" is construed.  In a software
context, that means that I disagree with the FSF's claim that linking
creates a derivative work.  In a book context, I don't believe that
accepting license under the GPL to copy Work X as a chapter or
appendix forces application of the GPL to the entire book.

As for the "extensive references", you can't claim license under the
GPL to include bits of Work X in an integral work of authorship that
is Chapter Y (or Chapters 1 through N) of your book, unless you are
willing to consider Chapter Y a derivative work and to release it
under the GPL.  However, if Chapters 1 through N are a separate work
of authorship (about which more later), combined with Work X to form a
compilation, they may well be defensible under a theory entirely
unrelated to the GPL, that of "fair use".

I will come back to "fair use" below, focusing first on whether the
GPL ought to be read to cover the whole book anyway.  The GPL
explicitly states in Section 0 that "a 'work based on the Program'
means either the Program or any derivative work under copyright law". 
It provides a paraphrase of that definition, which includes the phrase
"work containing the Program", but to read this automatically to mean
the biggest possible "work" (i. e., the whole book, or the whole boxed
set of which it is one volume) is contrary to its meaning in copyright

The term "derivative work" is defined in the Berne Convention, to
which the great majority of nations are signatories.  It appears in
that document once, as the title of Article 2, Clause 3, which says: 
"Translations, adaptations, arrangements of music and other
alterations of a literary or artistic work shall be protected as
original works without prejudice to the copyright in the original
work."  Articles 8 and 12 reserve to the author (or copyright holder)
the right to authorize, respectively, the creation of translations and
of the other listed forms of derivative work.

"Collections of literary or artistic works such as encyclopaedias and
anthologies" are discussed separately in Article 2, Clause 5, and are
considered a separate category of "intellectual creations".  It
appears clear to me that they are not intended to be included in the
category of "derivative works", but rather as a special case in which
the only protectable element belonging to the maker of the collection
is their "selection and arrangement".

Unlike all of the listed forms of derivative work, collections are not
mentioned in articles 8 and 12; the only other use of the word
"collection" in the text of the convention is in Article 2bis, Clause
3, which states essentially that individual works placed into the
public domain by their political or legal nature can still be selected
and arranged to form a copyrightable anthology.

Any given country's implementation of the Berne Convention may vary
somewhat, but the US statute (at least as of 1986) and the case law I
have seen are consistent with the interpretation that "compilations"
(or the subset "collective works") are a disjoint category from
"derivative works".  See 17 USC 101, and compare the UK CDPA sections
3 (1) (a) (a "table" or "compilation" is a subtype of literary work)
and 21 ("adaptations" are defined to cover essentially the same scope
as Berne Convention "derivative works").

OK, now on to the text of the GPL.  The "mere aggregation" clause that
ends Section 2 of the GPL purports to be an exception to the
requirement that larger works be entirely GPL.  In fact, the preceding
paragraph contains the only mention of "collective works" in the GPL,
stating that "the intent is to exercise the right to control the
distribution of derivative or collective works based on the Program". 
The paragraph before that, however, is worth quoting in its entirety:

"These requirements apply to the modified work as a whole. If
identifiable sections of that work are not derived from the Program,
and can be reasonably considered independent and separate works in
themselves, then this License, and its terms, do not apply to those
sections when you distribute them as separate works. But when you
distribute the same sections as part of a whole which is a work based
on the Program, the distribution of the whole must be on the terms of
this License, whose permissions for other licensees extend to the
entire whole, and thus to each and every part regardless of who wrote

Remember that Section 0 defined a "work based on the Program" to be a
derivative work, no more and no less.  Section 2 also begins with
language addressing modification to form a "work based on the
Program".  The above paragraph again talks about "a whole which is a
work based on the Program".  It's a close call, but I think it is
correct to read the entirety of Section 2 as being applicable only to
derivative works, and hence to permit Joe-Bob to accept license under
the GPL to copy Work X into his compilation, together with another
work that itself is legitimate under a "fair use" defense.

It's important to note that what we are construing here is not a
statute, it is an offer of contract.  A US court is obliged, as a
matter of law, to construe a statute in order to give each of its
provisions force; if this rule were applicable, it might be necessary
to rule that "work based on the Program" was meant to include
collective works all along, since otherwise the "mere aggregation"
clause would have little or no effect.

A contract, on the other hand, must as a matter of law be construed
against the offeror, and that goes double for a "standard-form"
contract in which there is no "meeting of the minds" (back-and-forth
negotiation) between offeror and offeree.  The GPL invites the reader
to interpret "derivative work" using its meaning under copyright law,
and thus to accept license to modify and copy Work X, maintaining its
status as a separately identifiable work of authorship, as part of a
compilation which also contains a critical work for which no license
is required.

As I said, this is a close call, and the need for the "fair use"
defense to justify the rest of Joe-Bob's book makes it closer in
practice.  Note also that the language of Section 2b, if read by
itself, does not use the phrase "work based on the Program" but rather
the more inclusive language "any work that you distribute or publish,
that in whole or in part contains or is derived from the Program or
any part thereof".  So there's some room for Sue to argue that Section
2b covers collective works even if nothing else in the GPL does; I
think that this argument should fail, but a court might not.

In the absence of the "mere aggregation" clause, the fact that Sue has
winked at the creation of other collective works containing Work X,
themselves not entirely covered by the GPL, could perhaps be used as a
wholly separate argument to estop her from pursuing this particular
use.  But in the presence of that clause, if Joe-Bob's argument that
the whole section only applies to derivative works were to fail, and
if the court held that Sue could reasonably have considered other uses
to have been "mere aggregation" while objecting to this use, then
Joe-Bob's conduct could be ruled to have breached the contract formed
under the GPL.

If Joe-Bob cannot rely on valid license under the GPL to include Work
X as Chapter (or Appendix) X of his book, then he has to fall back on
the argument that inclusion of the entire text of Work X is still
"fair use" without need for license.  This is now a pure copyright
infringement case and the GPL is out of the picture except as evidence
of the limited commercial value of non-GPL licenses to Work X as a
standalone work.

Joe-Bob might still win, since the most important "fair use" factor is
in his favor; the only effect his use could be said to have had on the
market for Work X is to reduce Sue's ability to charge for authorizing
a similar book by another author, and that author could write and
publish that book anyway, printing Work X as a separate volume.  But
Joe-Bob's publisher might be more prudent to create this separation in
the first place.

If GPL section 2b is to be read broadly, the borderline of "mere
aggregation" is anybody's guess.  Is it OK to put a Work X pamphlet in
a sleeve inside the book's back cover?  Is it OK to have typographical
uniformity but to ship the pamphlet to retailers under separate cover,
or is that some kind of contributory infringement because the retailer
or end user creates a "work" by combining them?  You wind up with the
same kind of hair-splitting that has occurred in the Eclipse+Kaffe

> 2. I fail to find the right technical or juridical terms here, but I
>    guess in most jurisdictions it is allowed to cite other texts, or to
>    publish a book that discusses some text in detail (like
>    interpretation of a poem, or detailed rebuttal of a scientific
>    paper). In such a case, the book would not exist without prior
>    existence of the original text. Would such a thing be regarded a
>    derived work, and would therefore a text published under GPL impose
>    restrictions that would not hold for a text published without a
>    license, simply in printed form?

As I mentioned above, the applicable defense here is "fair use", and
does not rely on any license terms but solely on statutory limits on
the copyright monopoly.  In the US at least, "fair use" operates as an
affirmative defense after it is established that substantive "copying"
of copyrightable fragments in excess of "de minimis" has taken place. 
The scope of "fair use" varies to some degree from jurisdiction to
jurisdiction, but is generally held to permit, for instance, critical
works that include fragments of an original work that are genuinely
necessary for coherent commentary.

Note that "fair use" is different from the argument that I have been
making elsewhere for the application-that-uses-a-GPL-library case,
which uses different precedents to arrive at the conclusion that the
bits of header file or equivalent published interface that wind up in
the application are uncopyrightable.  These precedents address the
functional aspect of the interface between a software component and
its user, the hardware/firmware on which it runs, or another software
component, and are only weakly applicable if at all to the case of
critical commentary on a text.

In the US, the criteria for "fair use" are defined in 17 USC 107 and
well illustrated by appellate precedents.  The criterion usually given
the most weight is "the effect of the use upon the potential market
for or value of the copyrighted work."  A critical essay is rarely
held to detract from the market for the original work, and is only
likely to fail as "fair use" if it incorporates the bulk of the
original without license.  Critical essays are routinely appended to
original works (published under license from the copyright holder) to
form critical editions without requiring additional license beyond
that for publication of the original.

So if 1) Joe-Bob has Sue's license under the GPL to distribute Work X,
2) the rest of Joe-Bob's book is defensible as "fair use" without any
need for license, and 3) it is correct to interpret the GPL to cover
Work X while not reaching out to grab the entire book (which may
depend on the details of bookbinding and distribution), then Joe-Bob
can successfully defend against Sue's claim of copyright infringement
and/or breach of contract.

Whew.  IANAL, etc., etc.

- Michael

Reply to: