Changeset - 764802727541
[Not reviewed]
0 2 0
Martin Michlmayr (tbm) - 10 years ago 2014-04-24 23:12:21
tbm@cyrius.com
Typo fixes
2 files changed with 3 insertions and 3 deletions:
0 comments (0 inline, 0 general)
compliance-guide.tex
Show inline comments
 
% compliance-guide.tex                            -*- LaTeX -*-
 

	
 
\part{A Practical Guide to GPL Compliance}
 
\label{gpl-compliance-guide}
 

	
 
{\parindent 0in
 
This part is: \\
 
\begin{tabbing}
 
Copyright \= \copyright{} 2014 \= \hspace{.2in} Bradley M. Kuhn. \\
 
Copyright \> \copyright{} 2008 \> \hspace{.2in} Software Freedom Law Center. \\
 
\end{tabbing}
 

	
 
\vspace{1in}
 

	
 
\begin{center}
 
Authors of this part are: \\
 

	
 
Bradley M. Kuhn \\
 
Aaron Williamson \\
 
Karen M. Sandler \\
 

	
 
\vspace{3in}
 

	
 
The copyright holders of this part hereby grant the freedom to copy, modify,
 
convey, Adapt, and/or redistribute this work under the terms of the Creative
 
Commons Attribution Share Alike 4.0 International License.  A copy of that
 
license is available at
 
\verb=https://creativecommons.org/licenses/by-sa/4.0/legalcode=. 
 
\end{center}
 
}
 

	
 
\bigskip
 

	
 
\chapter*{Executive Summary}
 

	
 
This is a guide to effective compliance with the GNU General Public
 
License (GPL) and related licenses.  Copyleft advocates
 
usually seek to assist the community with
 
GPL compliance cooperatively.   This guide focuses on complying from the
 
start, so that readers can learn to avoid enforcement actions entirely, or, at
 
least, minimize  the negative impact when enforcement actions occur.
 
This guide  introduces and explains basic legal concepts related to the GPL and its
 
enforcement by copyright holders. It also outlines business practices and
 
methods that lead to better GPL compliance.  Finally, it recommends proper
 
post-violation responses to the concerns of copyright holders.
 

	
 
\chapter{Background}
 

	
 
Early GPL enforcement efforts began soon after the GPL was written by
 
Richard M.~Stallman (RMS) in 1989, and consisted of informal community efforts,
 
often in public Usenet discussions.\footnote{One example is the public
 
  outcry over NeXT's attempt to make the Objective-C front-end to GCC
 
  proprietary.  RMS, in fact, handled this enforcement action personally and
 
  the Objective-C front-end is still part of upstream GCC today.}  Over the next decade, the Free Software Foundation (FSF),
 
which holds copyrights in many GNU programs, was the only visible entity
 
actively enforcing its GPL'd copyrights on behalf of the software freedom
 
community.
 
FSF's enforcement
 
was generally a private process; the FSF contacted violators
 
confidentially and helped them to comply with the license.  Most
 
violations were pursued this way until the early 2000's.
 

	
 
By that time, Linux-based systems such as GNU/Linux and BusyBox/Linux had become very common, particularly in
 
embedded devices such as wireless routers.  During this period, public
 
ridicule of violators in the press and on Internet fora supplemented
 
ongoing private enforcement and increased pressure on businesses to
 
comply.  In 2003, the FSF formalized its efforts into the GPL Compliance
 
Lab, increased the volume of enforcement, and built community coalitions
 
to encourage copyright holders to together settle amicably with violators.
 
Beginning in 2004, Harald Welte took a more organized public enforcement
 
approach and launched \verb0gpl-violations.org0, a website and mailing
 
list for collecting reports of GPL violations.  On the basis of these
 
reports, Welte successfully pursued many enforcements in Europe, including
 
formal legal action.  Harald earns the permanent fame as the first copyright
 
holder to bring legal action in a Court regarding GPL compliance. 
 

	
 
In 2007, two copyright holders in BusyBox, in conjunction with the
 
Software Freedom Conservancy (``Conservancy''), filed the first copyright infringement lawsuit
 
based on a violation of the GPL\@ in the USA. While  lawsuits are of course
 
quite public, the vast majority of Conservancy's enforcement actions 
 
are resolved privately via
 
cooperative communications with violators.  As both FSF and Conservancy has worked to bring
 
individual companies into compliance, both organizations have encountered numerous
 
violations resulting from preventable problems such as inadequate
 
attention to licensing of upstream software, misconceptions about the
 
GPL's terms, and poor communication between software developers and their
 
management.  This document highlights these problems and describe
 
best practices to encourage corporate Free Software users to reevaluate their
 
approach to GPL'd software and avoid future violations.
 

	
 
Both FSF and Conservancy continue GPL enforcement and compliance efforts
 
for software under the GPL, the GNU Lesser
 
Public License (LGPL) and other copyleft licenses.  In doing so, both organizations have
 
found that most violations stem from a few common mistakes that can be,
 
for the most part, easily avoided.  All copyleft advocates  hope to educate the community of
 
commercial distributors, redistributors, and resellers on how to avoid
 
violations in the first place, and to respond adequately and appropriately
 
when a violation occurs.
 

	
 
\chapter{Best Practices to Avoid Common Violations}
 
\label{best-practices}
 

	
 
Unlike highly permissive licenses (such as the ISC license), which
 
typically only require preservation of copyright notices, licensees face many
 
important requirements from the GPL.  These requirements are
 
carefully designed to uphold certain values and standards of the software
 
freedom community.  While the GPL's requirements may appear initially
 
counter-intuitive to those more familiar with proprietary software
 
licenses, by comparison, its terms are in fact clear and quite favorable to
 
licensees.  Indeed, the GPL's terms actually simplify compliance when
 
violations occur.
 

	
 
GPL violations occur (or, are compounded) most often when companies lack sound
 
practices for the incorporation of GPL'd components into their
 
internal development environment.  This section introduces some best
 
practices for software tool selection, integration and distribution,
 
inspired by and congruent with software freedom methodologies.  Companies should
 
establish such practices before building a product based on GPL'd
 
software.\footnote{This document addresses compliance with GPLv2,
 
  GPLv3, LGPLv2, and LGPLv3.  Advice on avoiding the most common
 
  errors differs little for compliance with these four licenses.
 
  \S~\ref{lgpl} discusses the key differences between GPL and LGPL
 
  compliance.}
 

	
 
\section{Evaluate License Applicability}
 
\label{derivative-works}
 
Political discussion about the GPL often centers around the ``copyleft''
 
requirements of the license.  Indeed, the license was designed primarily
 
to embody this licensing feature.  Most companies adding non-trivial
 
features (beyond mere porting and bug-fixing) to GPL'd software (and
 
thereby invoking these requirements) are already well aware of their
 
more complex obligations under the license.\footnote{While, there has been much legal
 
  discussion regarding copyleft and derivative works.  In practical
 
  reality, this issue is not relevant to the vast majority of companies
 
  distributing GPL'd software.  Those interested in this issue should study
 
  \tutorialpartsplit{\textit{Detailed Analysis of the GNU GPL and Related
 
      Licenses}'s Section on derivative works}{\S~\ref{derivative-works} of
 
    this tutorial}.}
 

	
 
However, experienced  GPL enforcers find that few redistributors'
 
compliance challenges relate directly to the copyleft provisions; this is
 
particularly true for most embedders.  Instead, the distributions of GPL'd
 
systems most often encountered typically consist of a full operating system
 
including components under the GPL (e.g., Linux, BusyBox) and components
 
under the LGPL (e.g., the GNU C Library).  Sometimes, these programs have
 
been patched or slightly improved by direct modification of their sources,
 
resulting unequivocally in a derivative work.  Alongside these programs,
 
companies often distribute fully independent, proprietary programs,
 
developed from scratch, which are designed to run on the Free Software operating
 
system but do not combine with, link to, modify, derive from, or otherwise
 
create a combined work with
 
the GPL'd components.\footnote{However, these programs do often combine
 
  with LGPL'd libraries. This is discussed in detail in \S~\ref{lgpl}.}
 
In the latter case, where the work is unquestionably a separate work of
 
creative expression, no copyleft provisions are invoked.
 

	
 
Admittedly, a tiny
 
minority of situations which lie outside these two categories, and thus
 
do involve close questions about derivative and combined works.  Those
 
situations admittedly do require a highly
 
fact-dependent analysis and cannot be addressed in a general-purpose
 
document, anyway.
 

	
 
\medskip
 

	
 
Most companies accused of violations lack a basic understanding
 
of how to comply even in the former straightforward scenario.  This document
 
provides those companies with the fundamental and generally applicable prerequisite knowledge.
 
For answers to rarer and more complicated legal questions, such as whether
 
your software is a derivative or combined work of some copylefted software, consult
 
with an attorney.\footnote{If you would like more information on the
 
  application of derivative works doctrine to software, a detailed legal
 
  discussion is presented in our colleague Dan Ravicher's article,
 
  \textit{Software Derivative Work: A Circuit Dependent Determination} and in
 
  \tutorialpartsplit{\textit{Detailed Analysis of the GNU GPL and Related
 
      Licenses}'s Section on derivative works}{\S~\ref{derivative-works} of
 
    this tutorial}.}
 

	
 
This discussion thus assumes that you have already identified the
 
``work'' covered by the license, and that any components not under the GPL
 
(e.g., applications written entirely by your developers that merely happen
 
to run on a Linux-based operating system) distributed in conjunction with
 
those works are separate works within the meaning of copyright law and the GPL\@.  In
 
such a case, the GPL requires you to provide complete corresponding
 
source (CCS)\footnote{For more on CCS,  see
 
\tutorialpartsplit{\textit{Detailed Analysis of the GNU GPL and Related
 
      Licenses}'s Section on GPLv2~\S2 and GPLv3~\S1.}{\S~\ref{GPLv2s2} and \S~\ref{GPLv3s1} of
 
    this tutorial}.}
 
for the GPL'd components and your modifications thereto, but not
 
for independent proprietary applications.  The procedures described in
 
this document address this typical scenario.
 

	
 
\section{Monitor Software Acquisition}
 

	
 
Software engineers deserve the freedom to innovate and import useful
 
software components to improve products.  However, along with that
 
freedom should come rules and reporting procedures to make sure that you
 
are aware of what software that you include with your product.
 

	
 
The most typical response to an initial enforcement action is: ``We
 
didn't know there was GPL'd stuff in there''.  This answer indicates
 
failure in the software acquisition and procurement process.  Integration
 
of third-party proprietary software typically requires a formal
 
arrangement and management/legal oversight before the developers
 
incorporate the software.  By contrast, developers often obtain and
 
integrate Free Software without intervention nor oversight. That ease of acquisition, however,
 
does not mean the oversight is any less necessary.  Just as your legal
 
and/or management team negotiates terms for inclusion of any proprietary
 
software, they should gently facilitate all decisions to bring Free Software into your
 
product.
 

	
 
Simple, engineering-oriented rules help provide a stable foundation for
 
Free Software integration.  For example, simply ask your software developers to send an email to a
 
standard place describing each new Free Software component they add to the system,
 
and have them include a brief description of how they will incorporate it
 
into the product.  Further, make sure developers use a revision control
 
system (such as Git or Mercurial), and have
 
store the upstream versions of all software in a ``vendor branch'' or
 
similar mechanism, whereby they can easily track and find the main version
 
of the software and, separately, any local changes.
 

	
 
Such procedures are best instituted at your project's launch.  Once 
 
chaotic and poorly-sourced development processes begin, cataloging the
 
presence of GPL'd components  becomes challenging.
 

	
 
Such a situation often requires use of a tool to ``catch up'' your knowledge
 
about what software your product includes.  Most commonly, companies choose
 
some software licensing scanning tool to inspect the codebase.  However,
 
there are few tools that are themselves Free Software.  Thus, GPL enforcers
 
usually recommend the GPL'd
 
\href{http://fossology.org/}{Fossology system}, which analyzes a
 
source code base and produces a list of Free Software licenses that may apply to
 
the code.  Fossology can help you build a catalog of the sources you have
 
already used to build your product.  You can then expand that into a more
 
structured inventory and process.
 

	
 
\section{Track Your Changes and Releases}
 

	
 
As explained in further detail below, the most important component of GPL
 
compliance is the one most often ignored: proper inclusion of CCS in all
 
distributions  of GPL'd
 
software.  To comply with GPL's CCS requirements, the distributor
 
\textit{must} always know precisely what sources generated a given binary
 
distribution.
 

	
 
In an unfortunately large number of our enforcement cases, the violating
 
company's engineering team had difficulty reconstructing the CCS
 
for binaries distributed by the company.  Here are three simple rules to
 
follow to decrease the likelihood of this occurance:
 
follow to decrease the likelihood of this occurrence:
 

	
 
\begin{itemize}
 

	
 
\item Ensure that your
 
developers are using revision control systems properly.
 

	
 
\item Have developers mark or ``tag'' the full source tree corresponding to
 
  builds distributed to customers
 

	
 
\item Check that your developers store all parts of the software
 
development in the revision control system, including {\sc readme}s, build
 
scripts, engineers' notes, and documentation.
 
\end{itemize}
 

	
 
Your developers will benefit anyway from these rules.  Developers will be
 
happier in their jobs if their tools already track the precise version of
 
source that corresponds to any deployed binary.
 

	
 
\section{Avoid the ``Build Guru''}
 

	
 
Too many software projects rely on only one or a very few team members who
 
know how to build and assemble the final released product.  Such knowledge
 
centralization not only creates engineering redundancy issues, but also
 
thwarts GPL compliance.  Specifically, CCS does not just require source code,
 
but scripts and other material that explain how to control compilation and
 
installation of the executable and object code.
 

	
 
Thus, avoid relying on a ``build guru'', a single developer who is the only one
 
who knows how to produce your final product. Make sure the build process
 
is well defined.  Train every developer on the build process for the final
 
binary distribution, including (in the case of embedded software)
 
generating a final firmware image suitable for distribution to the
 
customer.  Require developers to use revision control for build processes.
 
Make a rule that adding new components to the system without adequate
 
build instructions (or better yet, scripts) is unacceptable engineering
 
practice.
 

	
 
\chapter{Details of Compliant Distribution}
 

	
 
This section explains the specific requirements placed upon
 
distributors of GPL'd software.  Note that this section refers heavily to
 
specific provisions and language in
 
\href{http://www.gnu.org/licenses/old-licenses/gpl-2.0.html#section3}{GPLv2}
 
and \href{http://www.fsf.org/licensing/licenses/gpl.html#section6}{GPLv3}.
 
It may be helpful to have a copy of each license open while reading this
 
section.
 

	
 
\section{Binary Distribution Permission}
 
\label{binary-distribution-permission}
 

	
 
% be careful below, you cannot refill the \if section, so don't refill
 
% this paragraph without care.
 

	
 
The various versions of the GPL are copyright licenses that grant
 
permission to make certain uses of software that are otherwise restricted
 
by copyright law.  This permission is conditioned upon compliance with the
 
GPL's requirements.
 

	
 
This section walks through the requirements (of both GPLv2 and GPLv3) that
 
apply when you distribute GPL'd programs in binary (i.e., executable or
 
object code) form, which is typical for embedded applications.  Because a
 
binary application derives from a program's original sources, you need
 
permission from the copyright holder to distribute it.  \S~3 of GPLv2 and
 
\S~6 of GPLv3 contain the permissions and conditions related to binary
 
distributions of GPL'd programs.\footnote{These sections cannot be fully
 
  understood in isolation; read the entire license thoroughly before
 
  focusing on any particular provision.  However, once you have read and
 
  understood the entire license, look to these sections to guide
 
  compliance for binary distributions.}
 

	
 
GPL's binary distribution sections offer a choice of compliance methods,
 
each of which we consider in turn.  Each option refers to the
 
``Corresponding Source'' code for the binary distribution, which includes
 
the source code from which the binary was produced.  This abbreviated and
 
simplified definition is sufficient for the binary distribution discussion
 
in this section, but you may wish to refer back to this section after
 
reading the thorough discussion of ``Corresponding Source'' that appears
 
in \S~\ref{corresponding-source}.
 

	
 
\subsection{Option (a): Source Alongside Binary}
 

	
 
GPLv2~\S~3(a) and v3~\S~6(a) embody the easiest option for providing
 
source code: including Corresponding Source with every binary
 
distribution.  While other options appear initially less onerous, this
 
option invariably minimizes potential compliance problems, because when
 
you distribute Corresponding Source with the binary, \emph{your GPL
 
  obligations are satisfied at the time of distribution}.  This is not
 
true of other options, and for this reason, we urge you to seriously
 
consider this option.  If you do not, you may extend the duration of your
 
obligations far beyond your last binary distribution.
 

	
 
Compliance under this option is straightforward.  If you ship a product
 
that includes binary copies of GPL'd software (e.g., in firmware, or on a
 
hard drive, CD, or other permanent storage medium), you can store the
 
Corresponding Source alongside the binaries.  Alternatively, you can
 
include the source on a CD or other removable storage medium in the box
 
containing the product.
 

	
 
GPLv2 refers to the various storage mechanisms as ``medi[a] customarily
 
used for software interchange''.  While the Internet has attained primacy
 
as a means of software distribution where super-fast Internet connections
 
are available, GPLv2 was written at a time when downloading software was
 
not practical (and was often impossible).  For much of the world, this
 
condition has not changed since GPLv2's publication, and the Internet
 
still cannot be considered ``a medium customary for software
 
interchange''.  GPLv3 clarifies this matter, requiring that source be
 
``fixed on a durable physical medium customarily used for software
 
interchange''.  This language affirms that option (a) requires binary
 
redistributors to provide source on a physical medium.
 

	
 
Please note that while selection of option (a) requires distribution on a
 
physical medium, voluntary distribution via the Internet is very useful.  This
 
is discussed in detail in \S~\ref{offer-with-internet}.
 

	
 
\subsection{Option (b): The Offer}
 
\label{offer-for-source}
 

	
 
Many distributors prefer to ship only an offer for source with the binary
 
distribution, rather than the complete source package.  This
 
option has value when the cost of source distribution is a true
 
per-unit cost.  For example, this option might be a good choice for
 
embedded products with permanent storage too small to fit the source, and
 
which are not otherwise shipped with a CD but \emph{are} shipped with a
 
manual or other printed material.
 

	
 
However, this option increases the duration of your obligations
 
dramatically.  An offer for source must be good for three full years from
 
your last binary distribution (under GPLv2), or your last binary or spare
 
part distribution (under GPLv3).  Your source code request and
 
provisioning system must be designed to last much longer than your product
 
life cycle.
 

	
 
In addition, if you are required to comply with the terms of GPLv2, you
 
{\bf cannot} use a network service to provide the source code.  For GPLv2,
 
the source code offer is fulfilled only with physical media.  This usually
 
means that you must continue to produce an up-to-date ``source code CD''
 
for years after the product's end-of-life.
 

	
 
\label{offer-with-internet}
 

	
 
Under GPLv2, it is acceptable and advisable for your offer for source code
 
to include an Internet link for downloadable source \emph{in addition} to
 
offering source on a physical medium.  This practice enables those with
 
fast network connections to get the source more quickly, and typically
 
decreases the number of physical media fulfillment requests.
 
(GPLv3~\S~6(b) permits provision of source with a public
 
network-accessible distribution only and no physical media.  We discuss
 
this in detail at the end of this section.)
 

	
 
The following is a suggested compliant offer for source under GPLv2 (and
 
is also acceptable for GPLv3) that you would include in your printed
 
materials accompanying each binary distribution:
 

	
 
\begin{quote}
 
The software included in this product contains copyrighted software that
 
is licensed under the GPL\@.  A copy of that license is included in this
 
document on page $X$\@.  You may obtain the complete Corresponding Source
 
code from us for a period of three years after our last shipment of this
 
product, which will be no earlier than 2011-08-01, by sending a money
 
order or check for \$5 to: \\
 
GPL Compliance Division \\
 
Our Company \\
 
Any Town, US 99999 \\
 
\\
 
Please write ``source for product $Y$'' in the memo line of your
 
payment.
 

	
 
You may also find a copy of the source at
 
\verb0http://www.example.com/sources/Y/0.
 

	
 
This offer is valid to anyone in receipt of this information.
 
\end{quote}
 

	
 
There are a few important details about this offer.  First, it requires a
 
copying fee.  GPLv2 permits ``a charge no more than your cost of
 
physically performing source distribution''.  This fee must be reasonable.
 
If your cost of copying and mailing a CD is more than around \$10, you
 
should perhaps find a cheaper CD stock and shipment method.  It is simply
 
not in your interest to try to overcharge the community.  Abuse of this
 
provision in order to make a for-profit enterprise of source code
 
provision will likely trigger enforcement action.
 

	
 
Second, note that the last line makes the offer valid to anyone who
 
requests the source.  This is because v2~\S~3(b) requires that offers be
 
``to give any third party'' a copy of the Corresponding Source.  GPLv3 has
 
a similar requirement, stating that an offer must be valid for ``anyone
 
who possesses the object code''.  These requirements indicated in
 
v2~\S~3(c) and v3~\S~6(c) are so that noncommercial redistributors may
 
pass these offers along with their distributions.  Therefore, the offers
 
must be valid not only to your customers, but also to anyone who received
 
a copy of the binaries from them.  Many distributors overlook this
 
requirement and assume that they are only required to fulfill a request
 
from their direct customers.
 

	
 
The option to provide an offer for source rather than direct source
 
distribution is a special benefit to companies equipped to handle a
 
fulfillment process.  GPLv2~\S~3(c) and GPLv3~\S~6(c) avoid burdening
 
noncommercial, occasional redistributors with fulfillment request
 
obligations by allowing them to pass along the offer for source as they
 
received it.
 

	
 
Note that commercial redistributors cannot avail themselves of the option
 
(c) exception, and so while your offer for source must be good to anyone
 
who receives the offer (under v2) or the object code (under v3), it
 
\emph{cannot} extinguish the obligations of anyone who commercially
 
redistributes your product.  The license terms apply to anyone who
 
distributes GPL'd software, regardless of whether they are the original
 
distributor.  Take the example of Vendor $V$, who develops a software
 
platform from GPL'd sources for use in embedded devices.  Manufacturer $M$
 
contracts with $V$ to install the software as firmware in $M$'s device.
 
$V$ provides the software to $M$, along with a compliant offer for source.
 
In this situation, $M$ cannot simply pass $V$'s offer for source along to
 
its customers.  $M$ also distributes the GPL'd software commercially, so
 
$M$ too must comply with the GPL and provide source (or $M$'s \emph{own}
 
offer for source) to $M$'s customers.
 

	
 
This situation illustrates that the offer for source is often a poor
 
choice for products that your customers will likely redistribute.  If you
 
include the source itself with the products, then your distribution to
 
your customers is compliant, and their (unmodified) distribution to their
 
customers is likewise compliant, because both include source.  If you
 
include only an offer for source, your distribution is compliant but your
 
customer's distribution does not ``inherit'' that compliance, because they
 
have not made their own offer to accompany their distribution.
 

	
 
The terms related to the offer for source are quite different if you
 
distribute under GPLv3.  Under v3, you may make source available only over
 
a network server, as long as it is available to the general public and
 
remains active for three years from the last distribution of your product
 
or related spare part.  Accordingly, you may satisfy your fulfillment
 
obligations via Internet-only distribution.  This makes the ``offer for
 
source'' option less troublesome for v3-only distributions, easing
 
compliance for commercial redistributors.  However, before you switch to a
 
purely Internet-based fulfillment process, you must first confirm that you
 
can actually distribute \emph{all} of the software under GPLv3.  Some
 
programs are indeed licensed under ``GPLv2, \emph{or any later version}''
 
(often abbreviated ``GPLv2-or-later'').  Such licensing gives you the
 
option to redistribute under GPLv3.  However, a few popular programs are
 
only licensed under GPLv2 and not ``or any later version''
 
(``GPLv2-only'').  You cannot provide only Internet-based source request
 
fulfillment for the latter programs.
 

	
 
If you determine that all GPL'd works in your whole product allow upgrade
 
to GPLv3 (or were already GPLv3'd to start), your offer for source may be
 
as simple as this:
 

	
 
\begin{quote}
 
The software included in this product contains copyrighted software that
 
is licensed under the GPLv3\@.  A copy of that license is included in this
 
document on page $X$\@.  You may obtain the complete Corresponding Source
 
code from us for a period of three years after our last shipment of this
 
product and/or spare parts therefor, which will be no earlier than
 
2011-08-01, on our website at
 
\verb0http://www.example.com/sources/productnum/0.
 
\end{quote}
 

	
 
\medskip
 

	
 
Under both GPLv2 and GPLv3, source offers must be accompanied by a copy of
 
the license itself, either electronically or in print, with every
 
distribution.
 
 
 
Finally, it is unacceptable to use option (b) merely because you do not have
 
Corresponding Source ready.  We find that some companies chose this option
 
because writing an offer is easy, but producing a source distribution as
 
an afterthought to a hasty development process is difficult.  The offer
 
for source does not exist as a stop-gap solution for companies rushing to
 
market with an out-of-compliance product.  If you ship an offer for source
 
with your product but cannot actually deliver \emph{immediately} on that
 
offer when your customers receive it, you should expect an enforcement
 
action.
 

	
 
\subsection{Option (c): Noncommercial Offers}
 

	
 
As discussed in the last section, GPLv2~\S~3(c) and GPLv3~\S~6(c) apply
 
only to noncommercial use.  These options are not available to businesses
 
distributing GPL'd software.  Consequently, companies who redistribute
 
software packaged for them by an upstream vendor cannot merely pass along
 
the offer they received from the vendor; they must provide their own offer
 
or corresponding source to their distributees.  We talk in detail about
 
upstream software providers in \S~\ref{upstream}.
 

	
 
\subsection{Option 6(d) in GPLv3: Internet Distribution}
 

	
 
Under GPLv2, your formal provisioning options for Corresponding Source
 
ended with \S~3(c).  But even under GPLv2, pure Internet source
 
distribution was a common practice and generally considered to be
 
compliant.  GPLv2 mentions Internet-only distribution almost as aside in
 
the language, in text at the end of the section after the three
 
provisioning options are listed.  To quote that part of GPLv2~\S~3:
 
\begin{quote}
 
If distribution of executable or object code is made by offering access to
 
copy from a designated place, then offering equivalent access to copy the
 
source code from the same place counts as distribution of the source code,
 
even though third parties are not compelled to copy the source along with
 
the object code.
 
\end{quote}
 

	
 
When that was written in 1991, Internet distribution of software was the
 
exception, not the rule.  Some FTP sites existed, but generally software
 
was sent on magnetic tape or CDs.  GPLv2 therefore mostly assumed that
 
binary distribution happened on some physical media.  By contrast,
 
GPLv3~\S~6(d) explicitly gives an option for this practice that the
 
community has historically considered GPLv2-compliant.
 

	
 
Thus, you may fulfill your source-provision obligations by providing the
 
source code in the same way and from the same location.  When exercising
 
this option, you are not obligated to ensure that users download the
 
source when they download the binary, and you may use separate servers as
 
needed to fulfill the requests as long as you make the source as
 
accessible as the binary.  However, you must ensure that users can easily
 
find the source code at the time they download the binary. GPLv3~\S~6(d)
 
thus clarifies a point that has caused confusion about source provision in
 
v2.  Indeed, many such important clarifications are included in v3 which
 
together provide a compelling reason for authors and redistributors alike
 
to adopt GPLv3.
 

	
 
\subsection{Option 6(e) in GPLv3: Software Torrents}
 

	
 
Peer-to-peer file sharing arose well after GPLv2 was written, and does not
 
easily fit any of the v2 source provision options.  GPLv3~\S~6(e)
 
addresses this issue, explicitly allowing for distribution of source and
 
binary together on a peer-to-peer file sharing network.  If you distribute
 
solely via peer-to-peer networks, you can exercise this option.  However,
 
peer-to-peer source distribution \emph{cannot} fulfill your source
 
provision obligations for non-peer-to-peer binary distributions.  Finally,
 
you should ensure that binaries and source are equally seeded upon initial
 
peer-to-peer distribution.
 

	
 
\section{Preparing Corresponding Source}
 
\label{corresponding-source}
 

	
 
Most enforcement cases involve companies that have unfortunately not
 
implemented procedures like our \S~\ref{best-practices} recommendations
 
and have no source distribution arranged at all.  These companies must
 
work backwards from a binary distribution to come into compliance.  Our
 
recommendations in \S~\ref{best-practices} are designed to make it easy to
 
construct a complete and Corresponding Source release from the outset.  If
 
you have followed those principles in your development, you can meet the
 
following requirements with ease.  If you have not, you may have
 
substantial reconstruction work to do.
 

	
 
\subsection{Assemble the Sources}
 

	
 
For every binary that you produce, you should collect and maintain a copy
 
of the sources from which it was built.  A large system, such as an
 
embedded firmware, will probably contain many GPL'd and LGPL'd components
 
for which you will have to provide source.  The binary distribution may
 
also contain proprietary components which are separate and independent
 
works that are covered by neither the GPL nor LGPL\@.
 

	
 
The best way to separate out your sources is to have a subdirectory for
 
each component in your system.  You can then easily mark some of them as
 
required for your Corresponding Source releases.  Collecting
 
subdirectories of GPL'd and LGPL'd components is the first step toward
 
preparing your release.
 

	
 
\subsection{Building the Sources}
 

	
 
Few distributors, particularly of embedded systems, take care to read the
 
actual definition of Corresponding Source in the GPL\@.  Consider
 
carefully the definition, from GPLv3:
 
\begin{quote}
 
  The ``Corresponding Source'' for a work in object code form means all
 
  the source code needed to generate, install, and (for an executable
 
  work) run the object code and to modify the work, including scripts to
 
  control those activities.
 
\end{quote}
 

	
 
and the definition from GPLv2:
 
\begin{quote}
 
The source code for a work means the preferred form of the work for making
 
modifications to it.  For an executable work, complete source code means
 
all the source code for all modules it contains, plus any associated
 
interface definition files, plus the scripts used to control compilation
 
and installation of the executable.
 
\end{quote}
 

	
 
Note that you must include ``scripts used to control compilation and
 
installation of the executable'' and/or anything ``needed to generate,
 
install, and (for an executable work) run the object code and to modify
 
the work, including scripts to control those activities''.  These phrases
 
are written to cover different types of build environments and systems.
 
Therefore, the details of what you need to provide with regard to scripts
 
and installation instructions vary depending on the software details.  You
 
must provide all information necessary such that someone generally skilled
 
with computer systems could produce a binary similar to the one provided.
 

	
 
Take as an example an embedded wireless device.  Usually, a company
 
distributes a firmware, which includes a binary copy of
 
Linux\footnote{``Linux'' refers only to the kernel, not the larger system
 
  as a whole.} and a filesystem.  That filesystem contains various binary
 
programs, including some GPL'd binaries, alongside some proprietary
 
binaries that are separate works (i.e., not derived from, nor based on
 
freely-licensed sources).  Consider what, in this case, constitutes adequate
 
``scripts to control compilation and installation'' or items ``needed to
 
generate, install and run'' the GPL'd programs.
 

	
 
Most importantly, you must provide some sort of roadmap that allows
 
technically sophisticated users to build your software.  This can be
 
complicated in an embedded environment.  If your developers use scripts to
 
control the entire compilation and installation procedure, then you can
 
simply provide those scripts to users along with the sources they act
 
upon.  Sometimes, however, scripts were never written (e.g., the
 
information on how to build the binaries is locked up in the mind of your
 
``build guru'').  In that case, we recommend that you write out build
 
instructions in a natural language as a detailed, step-by-step {\sc
 
  readme}.
 

	
 
No matter what you offer, you need to give those who receive source a
 
clear path from your sources to binaries similar to the ones you ship.  If
 
you ship a firmware (kernel plus filesystem), and the filesystem contains
 
binaries of GPL'd programs, then you should provide whatever is necessary
 
to enable a reasonably skilled user to build any given GPL'd source
 
program (and modified versions thereof), and replace the given binary in
 
your filesystem.  If the kernel is Linux, then the users must have the
 
instructions to do the same with the kernel.  The best way to achieve this
 
is to make available to your users whatever scripts or process your
 
engineers would use to do the same.
 

	
 
These are the general details for how installation instructions work.
 
Details about what differs when the work is licensed under LGPL is
 
discussed in \S~\ref{lgpl}, and specific details that are unique to
 
GPLv3's installation instructions are in \S~\ref{user-products}.
 

	
 
\subsection{What About the Compiler?}
 

	
 
The GPL contains no provision that requires distribution of the compiler
 
used to build the software.  While companies are encouraged to make it as
 
easy as possible for their users to build the sources, inclusion of the
 
compiler itself is not normally considered mandatory.  The Corresponding
 
Source definition -- both in GPLv2 and GPLv3 -- has not been typically
 
read to include the compiler itself, but rather things like makefiles,
 
build scripts, and packaging scripts.
 

	
 
Nonetheless, in the interest of goodwill and the spirit of the GPL, most
 
companies do provide the compiler itself when they are able, particularly
 
when the compiler is based on GCC\@ or another copylefted compiler.  If you have
 
a GCC-based system, it is your prerogative to redistribute that GCC
 
version (binaries plus sources) to your customers.  We in the software freedom
 
community encourage you to do this, since it often makes it easier for
 
users to exercise their software freedom.  However, if you chose to take
 
this recommendation, ensure that your GCC distribution is itself
 
compliant.
 

	
 
If you have used a proprietary, third-party compiler to build the
 
software, then you probably cannot ship it to your customers.  We consider
 
the name of the compiler, its exact version number, and where it can be
 
acquired as information that \emph{must} be provided as part of the
 
Corresponding Source.  This information is essential to anyone who wishes
 
to produce a binary.  It is not the intent of the GPL to require you to
 
distribute third-party software tools to your customer (provided the tools
 
themselves are not based on the GPL'd software shipped), but we do believe
 
it requires that you give the user all the essential non-proprietary facts
 
that you had at your disposal to build the software.  Therefore, if you
 
choose not to distribute the compiler, you should include a {\sc readme}
 
about where you got it, what version it was, and who to contact to acquire
 
it, regardless of whether your compiler is Free Software, proprietary, or
 
internally developed.
 

	
 
\section{Best Practices and Corresponding Source}
 

	
 
\S~\ref{best-practices} and \S~\ref{corresponding-source} above are
 
closely related.  If you follow the best practices outlined above, you
 
will find that preparing your Corresponding Source release is an easier
 
task, perhaps even a trivial one.
 

	
 
Indeed, the enforcement process itself has historically been useful to
 
software development teams.  Development on a deadline can lead
 
organizations to cut corners in a way that negatively impacts its
 
development processes.  We have frequently been told by violators that
 
they experience difficulty when determining the exact source for a binary
 
in production (in some cases because their ``build guru'' quit during the
 
release cycle).  When management rushes a development team to ship a
 
release, they are less likely to keep release sources tagged and build
 
systems well documented.
 

	
 
We suggest that, if contacted about a violation, product builders use GPL
 
enforcement as an opportunity to improve their development practices.  No
 
developer would argue that their system is better for having a mysterious
 
build system and no source tracking.  Address these issues by installing a
 
revision system, telling your developers to use it, and requiring your
 
build guru to document his or her work!
 

	
 
\chapter{When The Letter Comes}
 

	
 
Unfortunately, many GPL violators ignore their obligations until they are
 
contacted by a copyright holder or the lawyer of a copyright holder.  You
 
should certainly contact your own lawyer if you have received a letter
 
alleging that you have infringed copyrights that were licensed to you
 
under the GPL\@.  This section outlines a typical enforcement case and
 
provides some guidelines for response.  These discussions are
 
generalizations and do not all apply to every alleged violation.
 

	
 
\section{Understanding Who's Enforcing}
 
\label{compliance-understanding-whos-enforcing}
 
% FIXME-LATER: this text needs work.
 

	
 
Both  FSF and Conservancy has, as part their mission,  to spread software
 
freedom. When FSF or Conservancy
 
enforces GPL, the goal is to bring the violator back into compliance as
 
quickly as possible, and redress the damage caused by the violation.
 
That is FSF's steadfast position in a violation negotiation --- comply
 
with the license and respect freedom.
 

	
 
However, other entities who do not share the full ethos of software freedom
 
as institutionalized by FSF pursue GPL violations differently.  Oracle, a
 
company that produces the GPL'd MySQL database, upon discovering GPL
 
violations typically negotiates a proprietary software license separately for
 
a fee.  While this practice is not one that FSF nor Conservancy would ever
 
consider undertaking or even endorsing, it is a legally way for copyright
 
holders to proceed.
 

	
 
Generally, GPL enforcers come in two varieties.  First, there are
 
Conservancy, FSF, and other ``community enforcers'', who primary seek the
 
policy goals of GPL (software freedom), and see financial compensation as
 
ultimately secondary to those goals.  Second, there are ``for-profit
 
enforcers'' who use the GPL as a either a crippleware license, or sneakily
 
induce infringement merely to gain proprietary licensing revenue.
 

	
 
Note that the latter model \textit{only} works for companies who hold 100\% of
 
the copyrights in the infringed work.  As such, multi-copyright-held works
 
are fully insulated from these tactics.
 

	
 

	
 
\section{Communication Is Key}
 

	
 
GPL violations are typically only escalated when a company ignores the
 
copyright holder's initial communication or fails to work toward timely
 
compliance.  Accused violators should respond very promptly to the
 
initial request.  As the process continues, violators should follow up weekly with the
 
copyright holders to make sure everyone agrees on targets and deadlines
 
for resolving the situation.
 

	
 
Ensure that any staff who might receive communications regarding alleged
 
GPL violations understands how to channel the communication appropriately
 
within your organization.  Often, initial contact is addressed for general
 
correspondence (e.g., by mail to corporate headquarters or by e-mail to
 
general informational or support-related addresses).  Train the staff that
 
processes such communications to escalate them to someone with authority
 
to take action.  An unknowledgable response to such an inquiry (e.g., from
 
a first-level technical support person) can cause negotiations to fail
 
prematurely.
 

	
 
Answer promptly by multiple means (paper letter, telephone call, and
 
email), even if your response merely notifies the sender that you are
 
investigating the situation and will respond by a certain date.  Do not
 
let the conversation lapse until the situation is fully resolved.
 
Proactively follow up with synchronous communication means to be sure
 
communications sent by non-reliable means (such as email) were received.
 

	
 
Remember that the software freedom community generally values open communication and
 
cooperation, and these values extend to GPL enforcement.  You will
 
generally find that software freedom developers and their lawyers are willing to
 
have a reasonable dialogue and will work with you to resolve a violation
 
once you open the channels of communication in a friendly way.
 

	
 
\section{Termination}
 

	
 
Many redistributors overlook the GPL's termination provision (GPLv2~\S~4 and
 
GPLv3~\S~8).  Under v2, violators forfeit their rights to redistribute and
 
modify the GPL'd software until those rights are explicitly reinstated by
 
the copyright holder.  In contrast, v3 allows violators to rapidly resolve
 
some violations without consequence.
 

	
 
If you have redistributed an application under GPLv2\footnote{This applies
 
  to all programs licensed to you under only GPLv2 (``GPLv2-only'').
 
  However, most so-called GPLv2 programs are actually distributed with
 
  permission to redistribute under GPLv2 \emph{or any later version of the
 
    GPL} (``GPLv2-or-later'').  In the latter cases, the redistributor can
 
  choose to redistribute under GPLv2, GPLv3, GPLv2-or-later or even
 
  GPLv3-or-later.  Where the redistributor has chosen v2 explicitly, the
 
  v2 termination provision will always apply.  If the redistributor has
 
  chosen v3, the v3 termination provision will always apply.  If the
 
  redistributor has chosen GPLv2-or-later, then the redistributor may want
 
  to narrow to GPLv3-only upon violation, to take advantage of the
 
  termination provisions in v3.}, but have violated the terms of GPLv2,
 
you must request a reinstatement of rights from the copyright holders
 
before making further distributions, or else cease distribution and
 
modification of the software forever.  Different copyright holders
 
condition reinstatement upon different requirements, and these
 
requirements can be (and often are) wholly independent of the GPL\@.  The
 
terms of your reinstatement will depend upon what you negotiate with the
 
copyright holder of the GPL'd program.
 

	
 
Since your rights under GPLv2 terminate automatically upon your initial
 
violation, \emph{all your subsequent distributions} are violations and
 
infringements of copyright.  Therefore, even if you resolve a violation on
 
your own, you must still seek a reinstatement of rights from the copyright
 
holders whose licenses you violated, lest you remain liable for
 
infringement for even compliant distributions made subsequent to the
 
initial violation.
 

	
 
GPLv3 is more lenient.  If you have distributed only v3-licensed programs,
 
you may be eligible under v3~\S~8 for automatic reinstatement of rights.
 
You are eligible for automatic reinstatement when:
 
\begin{itemize}
 
\item you correct the violation and are not contacted by a copyright
 
  holder about the violation within sixty days after the correction, or
 

	
 
\item you receive, from a copyright holder, your first-ever contact
 
  regarding a GPL violation, and you correct that violation within thirty
 
  days of receipt of copyright holder's notice.
 
\end{itemize}
 

	
 
In addition to these permanent reinstatements provided under v3, violators
 
who voluntarily correct their violation also receive provisional
 
permission to continue distributing until they receive contact from the
 
copyright holder.  If sixty days pass without contact, that reinstatement
 
becomes permanent.  Nonetheless, you should be prepared to cease
 
distribution during those initial sixty days should you receive a
 
termination notice from the copyright holder.
 

	
 
Given that much discussion of v3 has focused on its so-called more
 
complicated requirements, it should be noted that v3 is, in this regard,
 
more favorable to violators than v2.
 

	
 
\chapter{Standard Requests}
 

	
 
As we noted above, different copyright holders have different requirements
 
for reinstating a violator's distribution rights.  Upon violation, you no
 
longer have a license under the GPL\@.  Copyright holders can therefore
 
set their own requirements outside the license before reinstatement of
 
rights.  We have collected below a list of reinstatement demands that
 
copyright holders often require.
 

	
 
\begin{itemize}
 

	
 
\item {\bf Compliance on all Free Software copyrights}.  Copyright holders of Free Software
 
  often want a company to demonstrate compliance for all GPL'd software in
 
  a distribution, not just their own.  A copyright holder may refuse to
 
  reinstate your right to distribute one program unless and until you
 
  comply with the licenses of all Free Software in your distribution.
 
 
 
\item {\bf Notification to past recipients}.  Users to whom you previously
 
  distributed non-compliant software should receive a communication
 
  (email, letter, bill insert, etc.) indicating the violation, describing
 
  their rights under the GPL, and informing them how to obtain a gratis source
 
  distribution.  If a customer list does not exist (such as in reseller
 
  situations), an alternative form of notice may be required (such as a
 
  magazine advertisement).
 

	
 
\item {\bf Appointment of a GPL Compliance Officer.}  The software freedom community
 
  values personal accountability when things go wrong.  Copyright holders
 
  often require that you name someone within the violating company
 
  officially responsible for Free Software license compliance, and that this
 
  individual serve as the key public contact for the community when
 
  compliance concerns arise.
 

	
 
\item {\bf Periodic Compliance Reports.}  Many copyright holders wish to
 
  monitor future compliance for some period of time after the violation.
 
  For some period, your company may be required to send regular reports on
 
  how many distributions of binary and source have occurred.
 
\end{itemize}
 

	
 
These are just a few possible requirements for reinstatement.  In the
 
context of a GPL violation, and particularly under v2's termination
 
provision, the copyright holder may have a range of requests in exchange
 
for reinstatement of rights.  These software developers are talented
 
professionals from whose work your company has benefited.  Indeed, you are
 
unlikely to find a better value or more generous license terms for similar
 
software elsewhere.  Treat the copyright holders with the same respect you
 
treat your corporate partners and collaborators.
 

	
 
\chapter{Special Topics in Compliance}
 

	
 
There are several other issues that are less common, but also relevant in
 
a GPL compliance situation.  To those who face them, they tend to be of
 
particular interest.
 

	
 
\section{LGPL Compliance}
 
\label{lgpl}
 

	
 
GPL compliance and LGPL compliance mostly involve the same issues.  As we
 
discussed in \S~\ref{derivative-works}, questions of modified versions of
 
software are highly fact-dependant and cannot be easily addressed in any
 
overview document.  The LGPL adds some additional complexity to the
 
analysis.  Namely, the various LGPL versions permit proprietary licensing
 
of certain types of modified versions.  These issues are well beyond the
 
scope of this document, but as a rule of thumb, once you have determined
 
(in accordance with LGPLv3) what part of the work is the ``Application''
 
and what portions of the source are ``Minimal Corresponding Source'', then
 
you can usually proceed to follow the GPL compliance rules that we
 
discussed, replacing our discussion of ``Corresponding Source'' with
 
``Minimal Corresponding Source''.
 

	
 
LGPL also requires that you provide a mechanism to combine the Application
 
with a modified version of the library, and outlines some options for
 
this.  Also, the license of the whole work must permit ``reverse
 
engineering for debugging such modifications'' to the library.  Therefore,
 
you should take care that the EULA used for the Application does not
 
contradict this permission.
 

	
 
\section{Upstream Providers}
 
\label{upstream}
 

	
 
With ever-increasing frequency, software development (particularly for
 
embedded devices) is outsourced to third parties.  If you rely on an
 
upstream provider for your software, note that you \emph{cannot ignore
 
  your GPL compliance requirements} simply because someone else packaged
 
the software that you distribute.  If you redistribute GPL'd software
 
(which you do, whenever you ship a device with your upstream's software in
 
it), you are bound by the terms of the GPL\@.  No distribution (including
 
redistribution) is permissible absent adherence to the license terms.
 

	
 
Therefore, you should introduce a due diligence process into your software
 
acquisition plans.  This is much like the software-oriented
 
recommendations we make in \S~\ref{best-practices}.  Implementing
 
practices to ensure that you are aware of what software is in your devices
 
can only improve your general business processes.  You should ask a clear
 
list of questions of all your upstream providers and make sure the answers
 
are complete and accurate.  The following are examples of questions you
 
should ask:
 
\begin{itemize}
 

	
 
\item What are all the licenses that cover the software in this device?
 

	
 
\item From which upstream vendors, be they companies or individuals, did
 
  \emph{you} receive your software from before distributing it to us?
 

	
 
\item What are your GPL compliance procedures?
 

	
 
\item If there is GPL'd software in your distribution, we will be
 
  redistributors of this GPL'd software.  What mechanisms do you have in
 
  place to aid us with compliance?
 

	
 
\item If we follow your recommended compliance procedures, will you
 
  formally indemnify us in case we are nonetheless found to be in
 
  violation of the GPL?
 

	
 
\end{itemize}
 

	
 
This last point is particularly important.  Many GPL enforcements are
 
escalated because of petty finger-pointing between the distributor and its
 
upstream.  In our experience, agreements regarding GPL compliance issues
 
and procedures are rarely negotiated up front.  However, when they are,
 
violations are resolved much more smoothly (at least from the point of
 
view of the redistributor).
 

	
 
Consider the cost of potential violations in your acquisition process.
 
Using Free Software allows software vendors to reduce costs significantly, but be
 
wary of vendors who have done so without regard for the licenses.  If your
 
vendor's costs seem ``too good to be true,'' you may ultimately bear the
 
burden of the vendor's inattention to GPL compliance.  Ask the right
 
questions, demand an account of your vendors' compliance procedures, and
 
seek indemnity from them.
 

	
 
\section{User Products and Installation Information}
 
\label{user-products}
 

	
 
GPLv3 requires you to provide ``Installation Information'' when v3
 
software is distributed in a ``User Product.''  During the drafting of v3,
 
the debate over this requirement was contentious.  However, the provision
 
as it appears in the final license is reasonable and easy to understand.
 

	
 
If you put GPLv3'd software into a User Product (as defined by the
 
license) and \emph{you} have the ability to install modified versions onto
 
that device, you must provide information that makes it possible for the
 
user to install functioning, modified versions of the software.  Note that
 
if no one, including you, can install a modified version, this provision
 
does not apply.  For example, if the software is burned onto an
 
non-field-upgradable ROM chip, and the only way that chip can be upgraded
 
is by producing a new one via a hardware factory process, then it is
 
acceptable that the users cannot electronically upgrade the software
 
themselves.
 

	
 
Furthermore, you are permitted to refuse support service, warranties, and
 
software updates to a user who has installed a modified version.  You may
 
even forbid network access to devices that behave out of specification due
 
to such modifications.  Indeed, this permission fits clearly with usual
 
industry practice.  While it is impossible to provide a device that is
 
completely unmodifiable\footnote{Consider that the iPhone, a device
 
  designed primarily to restrict users' freedom to modify it, was unlocked
 
  and modified within 48 hours of its release.}, users are generally on
 
notice that they risk voiding their warranties and losing their update and
 
support services when they make modifications.\footnote{A popular t-shirt
 
  in the software freedom community reads: ``I void warranties.''.  Our community is
 
  well-known for modifying products with full knowledge of the
 
  consequences.  GPLv3's ``Installation Instructions'' section merely
 
  confirms that reality, and makes sure GPL rights can be fully exercised,
 
  even if users exercise those rights at their own peril.}
 

	
 
GPLv3 is in many ways better for distributors who seek some degree of
 
device lock-down.  Technical processes are always found for subverting any
 
lock-down; pursuing it is a losing battle regardless.  With GPLv3, unlike
 
with GPLv2, the license gives you clear provisions that you can rely on
 
when you are forced to cut off support, service or warranty for a customer
 
who has chosen to modify.
 

	
 
\chapter{Conclusion}
 

	
 
GPL compliance need not be an onerous process.  Historically, struggles
 
have been the result of poor development methodologies and communications,
 
rather than any unexpected application of the GPL's source code disclosure
 
requirements.
 

	
 
Compliance is straightforward when the entirety of your enterprise is
 
well-informed and well-coordinated.  The receptionists should know how to
 
route a GPL source request or accusation of infringement.  The lawyers
 
should know the basic provisions of Free Software licenses and your source
 
disclosure requirements, and should explain those details to the software
 
developers.  The software developers should use a version control system
 
that allows them to associate versions of source with distributed
 
binaries, have a well-documented build process that anyone skilled in the
 
art can understand, and inform the lawyers when they bring in new
 
software.  Managers should build systems and procedures that keep everyone
 
on target.  With these practices in place, any organization can comply
 
with the GPL without serious effort, and receive the substantial benefits
 
of good citizenship in the software freedom community, and lots of great code
 
ready-made for their products.
 

	
 
\vfill
 

	
 
% LocalWords:  redistributors NeXT's Slashdot Welte gpl ISC embedders BusyBox
 
% LocalWords:  someone's downloadable subdirectory subdirectories filesystem
 
% LocalWords:  roadmap README upstream's Ravicher's Fossology readme CDs iPhone
 
% LocalWords:  makefiles violator's
gpl-lgpl.tex
Show inline comments
...
 
@@ -1108,3254 +1108,3254 @@ commercial freedom.  GPLv2~\S1 clearly states that in the case of verbatim
 
copies, one may make money.  Re-distributors are fully permitted to charge
 
for the re-distribution of copies of Free Software. In addition, they may
 
provide the warranty protection that the GPL disclaims as an additional
 
service for a fee. (See Section~\ref{Business Models} for more discussion
 
on making a profit from Free Software redistribution.)
 

	
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 

	
 
\chapter{Derivative Works: Statute and Case Law}
 
\label{derivative-works}
 

	
 
We digress for this chapter from our discussion of GPL's exact text to
 
consider the matter of derivative works --- a concept that we must
 
understand fully before considering GPLv2~\S\S2--3\@. The GPL, and Free
 
Software licensing in general, relies critically on the concept of
 
``derivative work'' since software that is ``independent,'' (i.e., not
 
``derivative'') of Free Software need not abide by the terms of the
 
applicable Free Software license. As much is required by \S~106 of the
 
Copyright Act, 17 U.S.C. \S~106 (2002), and admitted by Free Software
 
licenses, such as the GPL, which (as we have seen) states in GPLv2~\S0 that ``a
 
`work based on the Program' means either the Program or any derivative
 
work under copyright law.'' It is being a derivative work of Free Software
 
that triggers the necessity to comply with the terms of the Free Software
 
license under which the original work is distributed. Therefore, one is
 
left to ask, just what is a ``derivative work''? The answer to that
 
question differs depending on which court is being asked.
 

	
 
The analysis in this chapter sets forth the differing definitions of
 
derivative work by the circuit courts. The broadest and most
 
established definition of derivative work for software is the
 
abstraction, filtration, and comparison test (``the AFC test'') as
 
created and developed by the Second Circuit. Some circuits, including
 
the Ninth Circuit and the First Circuit, have either adopted narrower
 
versions of the AFC test or have expressly rejected the AFC test in
 
favor of a narrower standard. Further, several other circuits have yet
 
to adopt any definition of derivative work for software.
 

	
 
As an introductory matter, it is important to note that literal copying of
 
a significant portion of source code is not always sufficient to establish
 
that a second work is a derivative work of an original
 
program. Conversely, a second work can be a derivative work of an original
 
program even though absolutely no copying of the literal source code of
 
the original program has been made. This is the case because copyright
 
protection does not always extend to all portions of a program's code,
 
while, at the same time, it can extend beyond the literal code of a
 
program to its non-literal aspects, such as its architecture, structure,
 
sequence, organization, operational modules, and computer-user interface.
 

	
 
\section{The Copyright Act}
 

	
 
The copyright act is of little, if any, help in determining the definition
 
of a derivative work of software. However, the applicable provisions do
 
provide some, albeit quite cursory, guidance. Section 101 of the Copyright
 
Act sets forth the following definitions:
 

	
 
\begin{quotation}
 
A ``computer program'' is a set of statements or instructions to be used
 
directly or indirectly in a computer in order to bring about a certain
 
result.
 

	
 
A ``derivative work'' is a work based upon one or more preexisting works,
 
such as a translation, musical arrangement, dramatization,
 
fictionalization, motion picture version, sound recording, art
 
reproduction, abridgment, condensation, or any other form in which a work
 
may be recast, transformed, or adapted. A work consisting of editorial
 
revisions, annotations, elaborations, or other modifications which, as a
 
whole, represent an original work of authorship, is a ``derivative work.''
 
\end{quotation}
 

	
 
These are the only provisions in the Copyright Act relevant to the
 
determination of what constitutes a derivative work of a computer
 
program. Another provision of the Copyright Act that is also relevant to
 
the definition of derivative work is \S~102(b), which reads as follows:
 

	
 
\begin{quotation}
 
In no case does copyright protection for an original work of authorship
 
extend to any idea, procedure, process, system, method of operation,
 
concept, principle, or discovery, regardless of the form in which it is
 
described, explained, illustrated, or embodied in such work.
 
\end{quotation}
 

	
 
Therefore, before a court can ask whether one program is a derivative work
 
of another program, it must be careful not to extend copyright protection
 
to any ideas, procedures, processes, systems, methods of operation,
 
concepts, principles, or discoveries contained in the original program. It
 
is the implementation of this requirement to ``strip out'' unprotectable
 
elements that serves as the most frequent issue over which courts
 
disagree.
 

	
 
\section{Abstraction, Filtration, Comparison Test}
 

	
 
As mentioned above, the AFC test for determining whether a computer
 
program is a derivative work of an earlier program was created by the
 
Second Circuit and has since been adopted in the Fifth, Tenth, and
 
Eleventh Circuits. Computer Associates Intl., Inc. v. Altai, Inc., 982
 
F.2d 693 (2nd Cir. 1992); Engineering Dynamics, Inc. v. Structural
 
Software, Inc., 26 F.3d 1335 (5th Cir. 1994); Kepner-Tregoe,
 
Inc. v. Leadership Software, Inc., 12 F.3d 527 (5th Cir. 1994); Gates
 
Rubber Co. v. Bando Chem. Indust., Ltd., 9 F.3d 823 (10th Cir. 1993);
 
Mitel, Inc. v. Iqtel, Inc., 124 F.3d 1366 (10th Cir. 1997); Bateman
 
v. Mnemonics, Inc., 79 F.3d 1532 (11th Cir. 1996); and, Mitek Holdings,
 
Inc. v. Arce Engineering Co., Inc., 89 F.3d 1548 (11th Cir. 1996).
 

	
 
Under the AFC test, a court first abstracts from the original program its
 
constituent structural parts. Then, the court filters from those
 
structural parts all unprotectable portions, including incorporated ideas,
 
expression that is necessarily incidental to those ideas, and elements
 
that are taken from the public domain. Finally, the court compares any and
 
all remaining kernels of creative expression to the structure of the
 
second program to determine whether the software programs at issue are
 
substantially similar so as to warrant a finding that one is the
 
derivative work of the other.
 

	
 
Often, the courts that apply the AFC test will perform a quick initial
 
comparison between the entirety of the two programs at issue in order to
 
help determine whether one is a derivative work of the other. Such a
 
holistic comparison, although not a substitute for the full application of
 
the AFC test, sometimes reveals a pattern of copying that is not otherwise
 
obvious from the application of the AFC test when, as discussed below,
 
only certain components of the original program are compared to the second
 
program. If such a pattern is revealed by the quick initial comparison,
 
the court is more likely to conclude that the second work is indeed a
 
derivative of the original.
 

	
 
\subsection{Abstraction}
 

	
 
The first step courts perform under the AFC test is separation of the
 
work's ideas from its expression. In a process akin to reverse
 
engineering, the courts dissect the original program to isolate each level
 
of abstraction contained within it. Courts have stated that the
 
abstractions step is particularly well suited for computer programs
 
because it breaks down software in a way that mirrors the way it is
 
typically created. However, the courts have also indicated that this step
 
of the AFC test requires substantial guidance from experts, because it is
 
extremely fact and situation specific.
 

	
 
By way of example, one set of abstraction levels is, in descending order
 
of generality, as follows: the main purpose, system architecture, abstract
 
data types, algorithms and data structures, source code, and object
 
code. As this set of abstraction levels shows, during the abstraction step
 
of the AFC test, the literal elements of the computer program, namely the
 
source and object code, are defined as particular levels of
 
abstraction. Further, the source and object code elements of a program are
 
not the only elements capable of forming the basis for a finding that a
 
second work is a derivative of the program. In some cases, in order to
 
avoid a lengthy factual inquiry by the court, the owner of the copyright in
 
the original work will submit its own list of what it believes to be the
 
protected elements of the original program. In those situations, the court
 
will forgo performing its own abstraction, and proceed to the second step of
 
the AFC test.
 

	
 
\subsection{Filtration}
 

	
 
The most difficult and controversial part of the AFC test is the second
 
step, which entails the filtration of protectable expression contained in
 
the original program from any unprotectable elements nestled therein. In
 
determining which elements of a program are unprotectable, courts employ a
 
myriad of rules and procedures to sift from a program all the portions
 
that are not eligible for copyright protection.
 

	
 
First, as set forth in \S~102(b) of the Copyright Act, any and all ideas
 
embodied in the program are to be denied copyright protection. However,
 
implementing this rule is not as easy as it first appears. The courts
 
readily recognize the intrinsic difficulty in distinguishing between ideas
 
and expression and that, given the varying nature of computer programs,
 
doing so will be done on an ad hoc basis. The first step of the AFC test,
 
the abstraction, exists precisely to assist in this endeavor by helping
 
the court separate out all the individual elements of the program so that
 
they can be independently analyzed for their expressive nature.
 

	
 
A second rule applied by the courts in performing the filtration step of
 
the AFC test is the doctrine of merger, which denies copyright protection
 
to expression necessarily incidental to the idea being expressed. The
 
reasoning behind this doctrine is that when there is only one way to
 
express an idea, the idea and the expression merge, meaning that the
 
expression cannot receive copyright protection due to the bar on copyright
 
protection extending to ideas. In applying this doctrine, a court will ask
 
whether the program's use of particular code or structure is necessary for
 
the efficient implementation of a certain function or process. If so, then
 
that particular code or structure is not protected by copyright and, as a
 
result, it is filtered away from the remaining protectable expression.
 

	
 
A third rule applied by the courts in performing the filtration step of
 
the AFC test is the doctrine of scenes a faire, which denies copyright
 
protection to elements of a computer program that are dictated by external
 
factors. Such external factors can include:
 

	
 
\begin{itemize}
 

	
 
  \item The mechanical
 
specifications of the computer on which a particular program is intended
 
to operate
 

	
 
  \item Compatibility requirements of other programs with which a
 
program is designed to operate in conjunction
 

	
 
  \item Computer manufacturers'
 
design standards
 

	
 
  \item Demands of the industry being serviced, and widely accepted programming practices within the computer industry
 

	
 
\end{itemize}
 

	
 
Any code or structure of a program that was shaped predominantly in
 
response to these factors is filtered out and not protected by
 
copyright. Lastly, elements of a computer program are also to be filtered
 
out if they were taken from the public domain or fail to have sufficient
 
originality to merit copyright protection.
 

	
 
Portions of the source or object code of a computer program are rarely
 
filtered out as unprotectable elements. However, some distinct parts of
 
source and object code have been found unprotectable. For example,
 
constants, the invariable integers comprising part of formulas used to
 
perform calculations in a program, are unprotectable. Further, although
 
common errors found in two programs can provide strong evidence of
 
copying, they are not afforded any copyright protection over and above the
 
protection given to the expression containing them.
 

	
 
\subsection{Comparison}
 

	
 
The third and final step of the AFC test entails a comparison of the
 
original program's remaining protectable expression to a second
 
program. The issue will be whether any of the protected expression is
 
copied in the second program and, if so, what relative importance the
 
copied portion has with respect to the original program overall. The
 
ultimate inquiry is whether there is ``substantial'' similarity between
 
the protected elements of the original program and the potentially
 
derivative work. The courts admit that this process is primarily
 
qualitative rather than quantitative and is performed on a case-by-case
 
basis. In essence, the comparison is an ad hoc determination of whether
 
the protectable elements of the original program that are contained in the
 
second work are significant or important parts of the original program. If
 
so, then the second work is a derivative work of the first. If, however,
 
the amount of protectable elements copied in the second work are so small
 
as to be de minimis, then the second work is not a derivative work of the
 
original.
 

	
 
\section{Analytic Dissection Test}
 

	
 
The Ninth Circuit has adopted the analytic dissection test to determine
 
whether one program is a derivative work of another. Apple Computer,
 
Inc. v. Microsoft Corp., 35 F.3d 1435 (9th Cir. 1994). The analytic
 
dissection test first considers whether there are substantial similarities
 
in both the ideas and expressions of the two works at issue. Once the
 
similar features are identified, analytic dissection is used to determine
 
whether any of those similar features are protected by copyright. This
 
step is the same as the filtration step in the AFC test. After identifying
 
the copyrightable similar features of the works, the court then decides
 
whether those features are entitled to ``broad'' or ``thin''
 
protection. ``Thin'' protection is given to non-copyrightable facts or
 
ideas that are combined in a way that affords copyright protection only
 
from their alignment and presentation, while ``broad'' protection is given
 
to copyrightable expression itself. Depending on the degree of protection
 
afforded, the court then sets the appropriate standard for a subjective
 
comparison of the works to determine whether, as a whole, they are
 
sufficiently similar to support a finding that one is a derivative work of
 
the other. ``Thin'' protection requires the second work be virtually
 
identical in order to be held a derivative work of an original, while
 
``broad'' protection requires only a ``substantial similarity.''
 

	
 
\section{No Protection for ``Methods of Operation''}
 

	
 
The First Circuit has taken the position that the AFC test is inapplicable 
 
when the works in question relate to unprotectable elements set forth in 
 
\S~102(b).  Their approach results in a much narrower definition
 
of derivative work for software in comparison to other circuits. Specifically, 
 
the
 
First Circuit holds that ``method of operation,'' as used in \S~102(b) of
 
the Copyright Act, refers to the means by which users operate
 
computers. Lotus Development Corp. v. Borland Int'l., Inc., 49 F.3d 807
 
(1st Cir. 1995).  In Lotus, the court held that a menu command
 
hierarchy for a computer program was uncopyrightable because it did not
 
merely explain and present the program's functional capabilities to the
 
user, but also served as a method by which the program was operated and
 
controlled. As a result, under the First Circuit's test, literal copying
 
of a menu command hierarchy, or any other ``method of operation,'' cannot
 
form the basis for a determination that one work is a derivative of
 
another.  As a result, courts in the First Circuit that apply the AFC test
 
do so only after applying a broad interpretation of \S~102(b) to filter out
 
unprotected elements. E.g., Real View, LLC v. 20-20 Technologies, Inc., 
 
683 F. Supp.2d 147, 154 (D. Mass. 2010).
 

	
 

	
 
\section{No Test Yet Adopted}
 

	
 
Several circuits, most notably the Fourth and Seventh, have yet to
 
declare their definition of derivative work and whether or not the
 
AFC, Analytic Dissection, or some other test best fits their
 
interpretation of copyright law. Therefore, uncertainty exists with
 
respect to determining the extent to which a software program is a
 
derivative work of another in those circuits. However, one may presume
 
that they would give deference to the AFC test since it is by far the
 
majority rule among those circuits that have a standard for defining
 
a software derivative work.
 

	
 
\section{Cases Applying Software Derivative Work Analysis}
 

	
 
In the preeminent case regarding the definition of a derivative work for
 
software, Computer Associates v. Altai, the plaintiff alleged that its
 
program, Adapter, which was used to handle the differences in operating
 
system calls and services, was infringed by the defendant's competitive
 
program, Oscar. About 30\% of Oscar was literally the same code as
 
that in Adapter. After the suit began, the defendant rewrote those
 
portions of Oscar that contained Adapter code in order to produce a new
 
version of Oscar that was functionally competitive with Adapter, without
 
having any literal copies of its code. Feeling slighted still, the
 
plaintiff alleged that even the second version of Oscar, despite having no
 
literally copied code, also infringed its copyrights. In addressing that
 
question, the Second Circuit promulgated the AFC test.
 

	
 
In abstracting the various levels of the program, the court noted a
 
similarity between the two programs' parameter lists and macros. However,
 
following the filtration step of the AFC test, only a handful of the lists
 
and macros were protectable under copyright law because they were either
 
in the public domain or required by functional demands on the
 
program. With respect to the handful of parameter lists and macros that
 
did qualify for copyright protection, after performing the comparison step
 
of the AFC test, it was reasonable for the district court to conclude that
 
they did not warrant a finding of infringement given their relatively minor
 
contribution to the program as a whole. Likewise, the similarity between
 
the organizational charts of the two programs was not substantial enough
 
to support a finding of infringement because they were too simple and
 
obvious to contain any original expression.
 

	
 
In the case of Oracle America v. Google, 872 F. Supp.2d 974 (N.D. Cal. 2012),
 
the Northern District of California District Court examined the question of 
 
whether the application program interfaces (APIs) associated with the Java
 
programming language are entitled to copyright protection.  While the 
 
court expressly declined to rule whether all APIs are free to use without 
 
license (872 F. Supp.2d 974 at 1002), the court held that the command 
 
structure and taxonomy of the APIs were not protectable under copyright law.
 
Specifically, the court characterized the command structure and taxonomy as
 
both a ``method of operation'' (using an approach not dissimilar to the 
 
First Circuit's analysis in Lotus) and a ``functional requirement for 
 
compatibility'' (using Sega v. Accolade, 977 F.2d 1510 (9th Cir. 1992) and
 
Sony Computer Ent. v. Connectix, 203 F.3d 596 (9th Cir. 2000) as analogies),
 
and thus unprotectable subject matter under \S~102(b). 
 

	
 
Perhaps not surprisingly, there have been few other cases involving a highly
 
detailed software derivative work analysis. Most often, cases involve
 
clearer basis for decision, including frequent bad faith on the part of
 
the defendant or over-aggressiveness on the part of the plaintiff.  
 

	
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 

	
 
\chapter{Modified Source and Binary Distribution}
 
\label{source-and-binary}
 

	
 
In this chapter, we discuss the two core sections that define the rights
 
and obligations for those who modify, improve, and/or redistribute GPL'd
 
software. These sections, GPLv2~\S\S2--3, define the central core rights and
 
requirements of GPLv2\@.
 

	
 
\section{GPLv2~\S2: Share and Share Alike}
 
\label{GPLv2s2}
 

	
 
For many, this is where the ``magic'' happens that defends software
 
freedom upon redistribution.  GPLv2~\S2 is the only place in GPLv2
 
that governs the modification controls of copyright law.  If users
 
modify a GPLv2'd program, they must follow the terms of GPLv2~\S2 in making
 
those changes.  Thus, this sections ensures that the body of GPL'd software, as it
 
continues and develops, remains Free as in freedom.
 

	
 
To achieve that goal, GPLv2~\S2 first sets forth that the rights of
 
redistribution of modified versions are the same as those for verbatim
 
copying, as presented in GPLv2~\S1.  Therefore, the details of charging money,
 
keeping copyright notices intact, and other GPLv2~\S1 provisions are in tact
 
here as well.  However, there are three additional requirements.
 

	
 
The first (GPLv2~\S2(a)) requires that modified files carry ``prominent
 
notices'' explaining what changes were made and the date of such
 
changes. This section does not prescribe some specific way of
 
marking changes nor does it control the process of how changes are made.
 
Primarily, GPLv2~\S2(a) seeks to ensure that those receiving modified
 
versions know the history of changes to the software.  For some users,
 
it is important to know that they are using the standard version of
 
program, because while there are many advantages to using a fork,
 
there are a few disadvantages.  Users should be informed about the
 
historical context of the software version they use, so that they can
 
make proper support choices.  Finally, GPLv2~\S2(a) serves an academic
 
purpose --- ensuring that future developers can use a diachronic
 
approach to understand the software.
 

	
 
\medskip
 

	
 
The second requirement (GPLv2~\S2(b)) contains the four short lines that embody
 
the legal details of ``share and share alike''.  These 46 words are
 
considered by some to be the most worthy of careful scrutiny because
 
GPLv2~\S2(b), and they
 
can be a source of great confusion when not properly understood.
 

	
 
In considering GPLv2~\S2(b), first note the qualifier: it \textit{only} applies to
 
derivative works that ``you distribute or publish''.  Despite years of
 
education efforts on this matter, many still believe that modifiers
 
of GPL'd software \textit{must} publish or otherwise
 
share their changes.  On the contrary, GPLv2~\S2(b) {\bf does not apply if} the
 
changes are never distributed.  Indeed, the freedom to make private,
 
personal, unshared changes to software for personal use only should be
 
protected and defended.\footnote{Most Free Software enthusiasts believe there is a {\bf
 
    moral} obligation to redistribute changes that are generally useful,
 
  and they often encourage companies and individuals to do so.  However, there
 
  is a clear distinction between what one {\bf ought} to do and what one
 
  {\bf must} do.}
 

	
 
Next, we again encounter the same matter that appears in GPLv2~\S0, in the
 
following text:
 
\begin{quote}
 
``...that in whole or part contains or is derived from the Program or any part thereof.''
 
\end{quote}
 
Again, the GPL relies here on what the copyright law says is a derivative
 
work.  If, under copyright law, the modified version ``contains or is
 
derived from'' the GPL'd software, then the requirements of GPLv2~\S2(b)
 
apply.  The GPL invokes its control as a copyright license over the
 
modification of the work in combination with its control over distribution
 
of the work.
 

	
 
The final clause of GPLv2~\S2(b) describes what the licensee must do if she is
 
distributing or publishing a work that is deemed a derivative work under
 
copyright law --- namely, the following:
 
\begin{quote}
 
[The work must] be licensed as a whole at no charge to all third parties
 
under the terms of this License.
 
\end{quote}
 
That is probably the most tightly-packed phrase in all of the GPL\@.
 
Consider each subpart carefully.
 

	
 
The work ``as a whole'' is what is to be licensed. This is an important
 
point that GPLv2~\S2 spends an entire paragraph explaining; thus this phrase is
 
worthy of a lengthy discussion here.  As a programmer modifies a software
 
program, she generates new copyrighted material --- fixing expressions of
 
ideas into the tangible medium of electronic file storage.  That
 
programmer is indeed the copyright holder of those new changes.  However,
 
those changes are part and parcel to the original work distributed to
 
the programmer under GPL\@. Thus, the license of the original work
 
affects the license of the new whole derivative work.
 

	
 
% {\cal I}
 
\newcommand{\gplusi}{$\mathcal{G\!\!+\!\!I}$}
 
\newcommand{\worki}{$\mathcal{I}$}
 
\newcommand{\workg}{$\mathcal{G}$}
 

	
 
\label{separate-and-independent}
 

	
 
It is certainly possible to take an existing independent work (called
 
\worki{}) and combine it with a GPL'd program (called \workg{}).  The
 
license of \worki{}, when it is distributed as a separate and independent
 
work, remains the prerogative of the copyright holder of \worki{}.
 
However, when \worki{} is combined with \workg{}, it produces a new work
 
that is the combination of the two (called \gplusi{}). The copyright of
 
this combined work, \gplusi{}, is held by the original copyright
 
holder of each of the two works.
 

	
 
In this case, GPLv2~\S2 lays out the terms by which \gplusi{} may be
 
distributed and copied.  By default, under copyright law, the copyright
 
holder of \worki{} would not have been permitted to distribute \gplusi{};
 
copyright law forbids it without the expressed permission of the copyright
 
holder of \workg{}. (Imagine, for a moment, if \workg{} were a proprietary
 
product --- would its copyright holders  give you permission to create and distribute
 
\gplusi{} without paying them a hefty sum?)  The license of \workg{}, the
 
GPL, states the  options for the copyright holder of \worki{}
 
who may want to create and distribute \gplusi{}. The  GPL's pre-granted
 
permission to create and distribute derivative works, provided the terms
 
of the GPL are upheld, goes far above and beyond the permissions that one
 
would get with a typical work not covered by a copyleft license.  (Thus, to
 
say that this condition is any way unreasonable is simply ludicrous.)
 

	
 
\medskip
 

	
 
\label{GPLv2s2-at-no-charge}
 
The next phrase of note in GPLv2~\S2(b) is ``licensed \ldots at no charge.''
 
This phrase  confuses many.  The sloppy reader points out this as ``a
 
contradiction in GPL'' because (in their confused view) that clause of GPLv2~\S2 says that re-distributors cannot
 
charge for modified versions of GPL'd software, but GPLv2~\S1 says that
 
they can.  Avoid this confusion: the ``at no charge'' \textbf{does not} prohibit re-distributors from
 
charging when performing the acts governed by copyright
 
law,\footnote{Recall that you could by default charge for any acts not
 
governed by copyright law, because the license controls are confined
 
by copyright.} but rather that they cannot charge a fee for the
 
\emph{license itself}.  In other words, redistributors of (modified
 
and unmodified) GPL'd works may charge any amount they choose for
 
performing the modifications on contract or the act of transferring
 
the copy to the customer, but they may not charge a separate licensing
 
fee for the software.
 

	
 
GPLv2~\S2(b) further states that the software must ``be licensed \ldots to all
 
third parties.''  This too yields some confusion, and feeds the
 
misconception mentioned earlier --- that all modified versions must be made
 
available to the public at large.  However, the text here does not say
 
that.  Instead, it says that the licensing under terms of the GPL must
 
extend to anyone who might, through the distribution chain, receive a copy
 
of the software.  Distribution to all third parties is not mandated here,
 
but GPLv2~\S2(b) does require re-distributors to license the derivative works in
 
a way that extends to all third parties who may ultimately receive a
 
copy of the software.
 

	
 
In summary, GPLv2\ 2(b) says what terms under which the third parties must
 
receive this no-charge license.  Namely, they receive it ``under the terms
 
of this License'', the GPLv2.  When an entity \emph{chooses} to redistribute
 
a derivative work of GPL'd software, the license of that whole 
 
work must be GPL and only GPL\@.  In this manner, GPLv2~\S2(b) dovetails nicely
 
with GPLv2~\S6 (as discussed in Section~\ref{GPLv2s6} of this tutorial).
 

	
 
\medskip
 

	
 
The final paragraph of GPLv2~\S2 is worth special mention.  It is possible and
 
quite common to aggregate various software programs together on one
 
distribution medium.  Computer manufacturers do this when they ship a
 
pre-installed hard drive, and GNU/Linux distribution vendors do this to
 
give a one-stop CD or URL for a complete operating system with necessary
 
applications.  The GPL very clearly permits such ``mere aggregation'' with
 
programs under any license.  Despite what you hear from its critics, the
 
GPL is nothing like a virus, not only because the GPL is good for you and
 
a virus is bad for you, but also because simple contact with a GPL'd
 
code-base does not impact the license of other programs.  A programmer must
 
expend actual effort  to cause a work to fall under the terms
 
of the GPL.  Redistributors are always welcome to simply ship GPL'd
 
software alongside proprietary software or other unrelated Free Software,
 
as long as the terms of GPL are adhered to for those packages that are
 
truly GPL'd.
 

	
 
\section{GPLv2~\S3: Producing Binaries}
 
\label{GPLv2s3}
 

	
 
Software is a strange beast when compared to other copyrightable works.
 
It is currently impossible to make a film or a book that can be truly
 
obscured.  Ultimately, the full text of a novel, even one written by
 
William Faulkner, must be presented to the reader as words in some
 
human-readable language so that they can enjoy the work.  A film, even one
 
directed by David Lynch, must be perceptible by human eyes and ears to
 
have any value.
 

	
 
Software is not so.  While the source code --- the human-readable
 
representation of software is of keen interest to programmers -- users and
 
programmers alike cannot make the proper use of software in that
 
human-readable form.  Binary code --- the ones and zeros that the computer
 
can understand --- must be predicable and attainable for the software to
 
be fully useful.  Without the binaries, be they in object or executable
 
form, the software serves only the didactic purposes of computer science.
 

	
 
Under copyright law, binary representations of the software are simply
 
derivative works of the source code.  Applying a systematic process (i.e.,
 
``compilation''\footnote{``Compilation'' in this context refers to the
 
  automated computing process of converting source code into binaries.  It
 
  has absolutely nothing to do with the term ``compilation'' in copyright statues.}) to a work of source code yields binary code. The binary
 
code is now a new work of expression fixed in the tangible medium of
 
electronic file storage.
 

	
 
Therefore, for GPL'd software to be useful, the GPL, since it governs the
 
rules for creation of derivative works, must grant permission for the
 
generation of binaries.  Furthermore, notwithstanding the relative
 
popularity of source-based GNU/Linux distributions like Gentoo, users find
 
it extremely convenient to receive distribution of binary software.  Such
 
distribution is the redistribution of derivative works of the software's
 
source code.  GPLv2~\S3 addresses the matter of creation and distribution of
 
binary versions.
 

	
 
Under GPLv2~\S3, binary versions may be created and distributed under the
 
terms of GPLv2~\S1--2, so all the material previously discussed applies
 
here.  However, GPLv2~\S3 must go a bit further.  Access to the software's
 
source code is an incontestable prerequisite for the exercise of the
 
fundamental freedoms to modify and improve the software.  Making even
 
the most trivial changes to a software program at the binary level is
 
effectively impossible.  GPLv2~\S3 must ensure that the binaries are never
 
distributed without the source code, so that these freedoms are passed
 
through the distribution chain.
 

	
 
GPLv2~\S3 permits distribution of binaries, and then offers three options for
 
distribution of source code along with binaries. The most common and the
 
least complicated is the option given under GPLv2~\S3(a).
 

	
 
GPLv2~\S3(a) offers the option to directly accompany the source code alongside
 
the distribution of the binaries.  This is by far the most convenient
 
option for most distributors, because it means that the source-code
 
provision obligations are fully completed at the time of binary
 
distribution (more on that later).
 

	
 
Under GPLv2~\S3(a), the source code provided must be the ``corresponding source
 
code.''  Here ``corresponding'' primarily means that the source code
 
provided must be that code used to produce the binaries being distributed.
 
That source code must also be ``complete''.   GPLv2~\S3's penultimate paragraph
 
explains in detail what is meant by ``complete''.  In essence, it is all
 
the material that a programmer of average skill would need to actually use
 
the source code to produce the binaries she has received.  Complete source
 
is required so that, if the licensee chooses, she should be able to
 
exercise her freedoms to modify and redistribute changes.  Without the
 
complete source, it would not be possible to make changes that were
 
actually directly derived from the version received.
 

	
 
Furthermore, GPLv2~\S3 is defending against a tactic that has in fact been
 
seen in GPL enforcement.  Under GPL, if you pay a high price for
 
a copy of GPL'd binaries (which comes with corresponding source, of
 
course), you have the freedom to redistribute that work at any fee you
 
choose, or not at all.  Sometimes, companies attempt a GPL-violating
 
cozenage whereby they produce very specialized binaries (perhaps for
 
an obscure architecture).  They then give source code that does
 
correspond, but withhold the ``incantations'' and build plans they
 
used to make that source compile into the specialized binaries.
 
Therefore, GPLv2~\S3 requires that the source code include ``meta-material'' like
 
scripts, interface definitions, and other material that is used to
 
``control compilation and installation'' of the binaries.  In this
 
manner, those further down the distribution chain are assured that
 
they have the unabated freedom to build their own derivative works
 
from the sources provided.
 

	
 
Software distribution comes in many
 
forms.  Embedded manufacturers, for example, have the freedom to put
 
GPL'd software into mobile devices with very tight memory and space
 
constraints.  In such cases, putting the source right alongside the
 
binaries on the machine itself might not be an option.  While it is
 
recommended that this be the default way that people comply with GPL, the
 
GPL does provide options when such distribution is unfeasible.
 

	
 
\label{GPLv2s3-medium-customarily}
 
GPLv2~\S3, therefore, allows source code to be provided on any physical
 
``medium customarily used for software interchange.''  By design, this
 
phrase covers a broad spectrum --- the phrase seeks to pre-adapt to
 
changes in  technology.  When GPLv2 was first published in June
 
1991, distribution on magnetic tape was still common, and CD was
 
relatively new.  By 2002, CD was the default.  By 2007, DVD's were the
 
default.  Now, it's common to give software on USB drives and SD cards.  This
 
language in the license must adapt with changing technology.
 

	
 
Meanwhile, the binding created by the word ``customarily'' is key.  Many
 
incorrectly believe that distributing binary on CD and source on the
 
Internet is acceptable.  In the corporate world in industrialized countries, it is indeed customary to
 
simply download a CDs' worth of data quickly.  However, even today in the USA, many computer users are not connected to the Internet, and most people connected
 
to the Internet still have limited download speeds.  Downloading
 
CDs full of data is not customary for them in the least.  In some cities
 
in Africa, computers are becoming more common, but Internet connectivity
 
is still available only at a few centralized locations.  Thus, the
 
``customs'' here are normalized for a worldwide userbase.  Simply
 
providing source on the Internet --- while it is a kind, friendly and
 
useful thing to do --- is not usually sufficient.
 

	
 
Note, however, a major exception to this rule, given by the last paragraph
 
of GPLv2~\S3. \emph{If} distribution of the binary files is made only on the
 
Internet (i.e., ``from a designated place''), \emph{then} simply providing
 
the source code right alongside the binaries in the same place is
 
sufficient to comply with GPLv2~\S3.
 

	
 
\medskip
 

	
 
As is shown above, under GPLv2~\S3(a), embedded manufacturers can put the
 
binaries on the device and ship the source code along on a CD\@.  However,
 
sometimes this turns out to be too costly.  Including a CD with every
 
device could prove too costly, and may practically (although not legally)
 
prohibit using GPL'd software. For this situation and others like it, GPLv2\S~3(b) is available.
 

	
 
GPLv2~\S3(b) allows a distributor of binaries to instead provide a written
 
offer for source code alongside those binaries.  This is useful in two
 
specific ways.  First, it may turn out that most users do not request the
 
source, and thus the cost of producing the CDs is saved --- a financial
 
and environmental windfall.  In addition, along with a GPLv2~\S3(b) compliant
 
offer for source, a binary distributor might choose to \emph{also} give a
 
URL for source code.  Many who would otherwise need a CD with source might
 
turn out to have those coveted high bandwidth connections, and are able to
 
download the source instead --- again yielding environmental and financial
 
windfalls.
 

	
 
However, note that regardless of how many users prefer to get the
 
source online, GPLv2~\S3(b) does place lasting long-term obligations on the
 
binary distributor.  The binary distributor must be prepared to honor
 
that offer for source for three years and ship it out (just as they
 
would have had to do under GPLv2~\S3(a)) at a moment's notice when they
 
receive such a request.  There is real organizational cost here:
 
support engineers must be trained how to route source requests, and
 
source CD images for every release version for the last three years
 
must be kept on hand to burn such CDs quickly. The requests might not
 
even come from actual customers; the offer for source must be valid
 
for ``any third party''.
 

	
 
That phrase is another place where some get confused --- thinking again
 
that full public distribution of source is required.  The offer for source
 
must be valid for ``any third party'' because of the freedoms of
 
redistribution granted by GPLv2~\S\S1--2.  A company may ship a binary image
 
and an offer for source to only one customer.  However, under GPL, that
 
customer has the right to redistribute that software to the world if she
 
likes.  When she does, that customer has an obligation to make sure that
 
those who receive the software from her can exercise their freedoms under
 
GPL --- including the freedom to modify, rebuild, and redistribute the
 
source code.
 

	
 
GPLv2~\S3(c) is created to save her some trouble, because by itself GPLv2~\S3(b)
 
would unfairly favor large companies.  GPLv2~\S3(b) allows the
 
separation of the binary software from the key tool that people can use
 
to exercise their freedom. The GPL permits this separation because it is
 
good for re-distributors, and those users who turn out not to need the
 
source.  However, to ensure equal rights for all software users, anyone
 
along the distribution chain must have the right to get the source and
 
exercise those freedoms that require it.
 

	
 
Meanwhile, GPLv2~\S3(b)'s compromise primarily benefits companies that
 
distribute binary software commercially.  Without GPLv2~\S3(c), that benefit
 
would be at the detriment of the companies' customers; the burden of
 
source code provision would be unfairly shifted to the companies'
 
customers.  A customer, who had received binaries with a GPLv2~\S3(b)-compliant
 
offer, would be required under GPLv2 (sans GPLv2~\S3(c)) to acquire the source,
 
merely to give a copy of the software to a friend who needed it.  GPLv2~\S3(c)
 
reshifts this burden to entity who benefits from GPLv2~\S3(b).
 

	
 
GPLv2~\S3(c) allows those who undertake \emph{noncommercial} distribution to
 
simply pass along a GPLv2~\S3(b)-compliant source code offer.  The customer who
 
wishes to give a copy to her friend can now do so without provisioning the
 
source, as long as she gives that offer to her friend.  By contrast, if
 
she wanted to go into business for herself selling CDs of that software,
 
she would have to acquire the source and either comply via GPLv2~\S3(a), or
 
write her own GPLv2~\S3(b)-compliant source offer.
 

	
 
This process is precisely the reason why a GPLv2~\S3(b) source offer must be
 
valid for all third parties.  At the time the offer is made, there is no
 
way of knowing who might end up noncommercially receiving a copy of the
 
software.  Companies who choose to comply via GPLv2~\S3(b) must thus be
 
prepared to honor all incoming source code requests.  For this and the
 
many other additional necessary complications under GPLv2~\S\S3(b--c), it is
 
only rarely a better option than complying via GPLv2~\S3(a).
 

	
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 
\chapter{GPL's Implied Patent Grant}
 
\label{gpl-implied-patent-grant}
 

	
 
We digress again briefly from our section-by-section consideration of GPLv2
 
to consider the interaction between the terms of GPL and patent law. The
 
GPLv2, despite being silent with respect to patents, actually confers on its
 
licensees more rights to a licensor's patents than those licenses that
 
purport to address the issue. This is the case because patent law, under
 
the doctrine of implied license, gives to each distributee of a patented
 
article a license from the distributor to practice any patent claims owned
 
or held by the distributor that cover the distributed article. The
 
implied license also extends to any patent claims owned or held by the
 
distributor that cover ``reasonably contemplated uses'' of the patented
 
article. To quote the Federal Circuit Court of Appeals, the highest court
 
for patent cases other than the Supreme Court:
 

	
 
\begin{quotation}
 
Generally, when a seller sells a product without restriction, it in
 
effect promises the purchaser that in exchange for the price paid, it will
 
not interfere with the purchaser's full enjoyment of the product
 
purchased. The buyer has an implied license under any patents of the
 
seller that dominate the product or any uses of the product to which the
 
parties might reasonably contemplate the product will be put.
 
\end{quotation}
 
Hewlett-Packard Co. v. Repeat-O-Type Stencil Mfg. Corp., Inc., 123 F.3d
 
1445, 1451 (Fed. Cir. 1997).
 

	
 
Of course, Free Software is licensed, not sold, and there are indeed
 
restrictions placed on the licensee, but those differences are not likely
 
to prevent the application of the implied license doctrine to Free
 
Software, because software licensed under the GPL grants the licensee the
 
right to make, use, and sell the software, each of which are exclusive
 
rights of a patent holder. Therefore, although the GPLv2 does not expressly
 
grant the licensee the right to do those things under any patents the
 
licensor may have that cover the software or its reasonably contemplated
 
uses, by licensing the software under the GPLv2, the distributor impliedly
 
licenses those patents to the GPLv2 licensee with respect to the GPLv2'd
 
software.
 

	
 
An interesting issue regarding this implied patent license of GPLv2'd
 
software is what would be considered ``uses of the [software] to which
 
the parties might reasonably contemplate the product will be put.'' A
 
clever advocate may argue that the implied license granted by GPLv2 is
 
larger in scope than the express license in other Free Software
 
licenses with express patent grants, in that the patent license
 
clause of many of those other Free  Software licenses are specifically 
 
limited to the patent claims covered by the code as licensed by the patentee.
 

	
 
In contrast, a GPLv2 licensee, under the doctrine of implied patent license, 
 
is free to practice any patent claims held by the licensor that cover 
 
``reasonably contemplated uses'' of the GPL'd code, which may very well 
 
include creation and distribution of derivative works since the GPL's terms, 
 
under which the patented code is distributed, expressly permits such activity.
 

	
 

	
 
Further supporting this result is the Federal Circuit's pronouncement that
 
the recipient of a patented article has, not only an implied license to
 
make, use, and sell the article, but also an implied patent license to
 
repair the article to enable it to function properly, Bottom Line Mgmt.,
 
Inc. v. Pan Man, Inc., 228 F.3d 1352 (Fed. Cir. 2000). Additionally, the
 
Federal Circuit extended that rule to include any future recipients of the
 
patented article, not just the direct recipient from the distributor.
 
This theory comports well with the idea of Free Software, whereby software
 
is distributed among many entities within the community for the purpose
 
of constant evolution and improvement. In this way, the law of implied
 
patent license used by the GPLv2 ensures that the community mutually
 
benefits from the licensing of patents to any single community member.
 

	
 
Note that simply because GPLv2'd software has an implied patent license does
 
not mean that any patents held by a distributor of GPLv2'd code become
 
worthless. To the contrary, the patents are still valid and enforceable
 
against either:
 

	
 
\begin{enumerate}
 
 \renewcommand{\theenumi}{\alph{enumi}}
 
 \renewcommand{\labelenumi}{\textup{(\theenumi)}}
 

	
 
\item any software other than that licensed under the GPLv2 by the patent
 
  holder, and
 

	
 
\item any party that does not comply with the GPLv2
 
with respect to the licensed software.
 
\end{enumerate}
 

	
 
\newcommand{\compB}{$\mathcal{B}$}
 
\newcommand{\compA}{$\mathcal{A}$}
 

	
 
For example, if Company \compA{} has a patent on advanced Web browsing, but
 
also licenses a Web browsing program under the GPLv2, then it
 
cannot assert the patent against any party based on that party's use of 
 
Company \compA{}'s GPL'ed Web browsing software program, or on that party's
 
creation and use of derivative works of that GPL'ed program.  However, if a
 
party uses that program without
 
complying with the GPLv2, then Company \compA{} can assert both copyright
 
infringement claims against the non-GPLv2-compliant party and
 
infringement of the patent, because the implied patent license only
 
extends to use of the software in accordance with the GPLv2. Further, if
 
Company \compB{} distributes a competitive advanced Web browsing program 
 
that is not a derivative work of Company \compA{}'s GPL'd Web browsing software
 
program, Company \compA{} is free to assert its patent against any user or
 
distributor of that product. It is irrelevant whether Company \compB's
 
program is also distributed under the GPLv2, as Company \compB{} can not grant
 
implied licenses to Company \compA's patent.
 

	
 
This result also reassures companies that they need not fear losing their
 
proprietary value in patents to competitors through the GPLv2 implied patent
 
license, as only those competitors who adopt and comply with the GPLv2's
 
terms can benefit from the implied patent license. To continue the
 
example above, Company \compB{} does not receive a free ride on Company
 
\compA's patent, as Company \compB{} has not licensed-in and then
 
redistributed Company A's advanced Web browser under the GPLv2. If Company
 
\compB{} does do that, however, Company \compA{} still has not lost
 
competitive advantage against Company \compB{}, as Company \compB{} must then,
 
when it re-distributes Company \compA's program, grant an implied license
 
to any of its patents that cover the program. Further, if Company \compB{}
 
relicenses an improved version of Company A's program, it must do so under
 
the GPLv2, meaning that any patents it holds that cover the improved version
 
are impliedly licensed to any licensee. As such, the only way Company
 
\compB{} can benefit from Company \compA's implied patent license, is if it,
 
itself, distributes Company \compA's software program and grants an
 
implied patent license to any of its patents that cover that program.
 

	
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 
\chapter{Defending Freedom on Many Fronts}
 

	
 
Chapters~\ref{run-and-verbatim} and~\ref{source-and-binary} presented the
 
core freedom-defending provisions of GPLv2\@, which are in GPLv2~\S\S0--3.
 
GPLv2\S\S~4--7 of the GPLv2 are designed to ensure that GPLv2~\S\S0--3 are
 
not infringed, are enforceable, are kept to the confines of copyright law but
 
also  not trumped by other copyright agreements or components of other
 
entirely separate legal systems.  In short, while GPLv2~\S\S0--3 are the parts
 
of the license that defend the freedoms of users and programmers,
 
GPLv2~\S\S4--7 are the parts of the license that keep the playing field clear
 
so that \S\S~0--3 can do their jobs.
 

	
 
\section{GPLv2~\S4: Termination on Violation}
 
\label{GPLv2s4}
 

	
 
GPLv2~\S4 is GPLv2's termination clause.  Upon first examination, it seems
 
strange that a license with the goal of defending users' and programmers'
 
freedoms for perpetuity in an irrevocable way would have such a clause.
 
However, upon further examination, the difference between irrevocability
 
and this termination clause becomes clear.
 

	
 
The GPL is irrevocable in the sense that once a copyright holder grants
 
rights for someone to copy, modify and redistribute the software under terms
 
of the GPL, they cannot later revoke that grant.  Since the GPL has no
 
provision allowing the copyright holder to take such a prerogative, the
 
license is granted as long as the copyright remains in effect.\footnote{In
 
  the USA, due to unfortunate legislation, the length of copyright is nearly
 
  perpetual, even though the Constitution forbids perpetual copyright.} The
 
copyright holders have the right to relicense the same work under different
 
licenses (see Section~\ref{Proprietary Relicensing} of this tutorial), or to
 
stop distributing the GPLv2'd version (assuming GPLv2~\S3(b) was never used),
 
but they may not revoke the rights under GPLv2 already granted.
 

	
 
In fact, when an entity loses their right to copy, modify and distribute
 
GPL'd software, it is because of their \emph{own actions}, not that of the
 
copyright holder.  The copyright holder does not decide when GPLv2~\S4
 
termination occurs (if ever); rather, the actions of the licensee determine
 
that.
 

	
 
Under copyright law, the GPL has granted various rights and freedoms to
 
the licensee to perform specific types of copying, modification, and
 
redistribution.  By default, all other types of copying, modification, and
 
redistribution are prohibited.  GPLv2~\S4 says that if you undertake any of
 
those other types (e.g., redistributing binary-only in violation of GPLv2~\S3),
 
then all rights under the license --- even those otherwise permitted for
 
those who have not violated --- terminate automatically.
 

	
 
GPLv2~\S4 makes GPLv2 enforceable.  If licensees fail to adhere to the
 
license, then they are stuck without any permission under to engage in
 
activities covered by copyright law.  They must completely cease and desist
 
from all copying, modification and distribution of the GPL'd software.
 

	
 
At that point, violating licensees must gain the forgiveness of the copyright
 
holders to have their rights restored.  Alternatively, the violators could
 
negotiate another agreement, separate from GPL, with the copyright
 
holder.  Both are common practice, although
 
\tutorialpartsplit{as discussed in \textit{A Practical Guide to GPL
 
    Compliance}, there are }{Chapter~\ref{compliance-understanding-whos-enforcing}
 
  explains further} key differences between these two very different uses of GPL.
 

	
 
\section{GPLv2~\S5: Acceptance, Copyright Style}
 
\label{GPLv2s5}
 

	
 
GPLv2~\S5 brings us to perhaps the most fundamental misconception and common
 
confusion about GPLv2\@. Because of the prevalence of proprietary software,
 
most users, programmers, and lawyers alike tend to be more familiar with
 
EULAs. EULAs are believed by their authors to be contracts, requiring
 
formal agreement between the licensee and the software distributor to be
 
valid. This has led to mechanisms like ``shrink-wrap'' and ``click-wrap''
 
as mechanisms to perform acceptance ceremonies with EULAs.
 

	
 
The GPL does not need contract law to ``transfer rights.''  Usually, no rights
 
are transferred between parties.  By contrast, the GPL is primarily a permission
 
slip to undertake activities that would otherwise have been prohibited
 
by copyright law.  As such, GPL needs no acceptance ceremony; the
 
licensee is not even required to accept the license.
 

	
 
However, without the GPL, the activities of copying, modifying and
 
distributing the software would have otherwise been prohibited.  So, the
 
GPL says that you only accepted the license by undertaking activities that
 
you would have otherwise been prohibited without your license under GPL\@.
 
This is a certainly subtle point, and requires a mindset quite different
 
from the contractual approach taken by EULA authors.
 

	
 
An interesting side benefit to GPLv2~\S5 is that the bulk of users of Free
 
Software are not required to accept the license.  Undertaking fair and
 
unregulated use of the work, for example, does not bind you to the GPL,
 
since you are not engaging in activity that is otherwise controlled by
 
copyright law.  Only when you engage in those activities that might have an
 
impact on the freedom of others does license acceptance occur, and the
 
terms begin to bind you to fair and equitable sharing of the software.  In
 
other words, the GPL only kicks in when it needs to for the sake of
 
freedom.
 

	
 
While GPL is by default a copyright license, it is certainly still possible
 
to consider GPL as a contract as well.  For example, some distributors chose
 
to ``wrap'' their software in an acceptance ceremony to the GPL, and nothing in
 
the GPL prohibits that use.  Furthermore, the ruling in \textit{Jacobsen
 
  v. Katzer, 535 F.3d 1373, 1380 (Fed.Cir.2008)} indicates that \textbf{both}
 
copyright and contractual remedies may be sought by a copyright holder
 
seeking to enforce a license designed to uphold software freedom.
 

	
 
% FIXME-LATER: Write this
 

	
 
%\section{Using GPL Both as a Contract and Copyright License}
 

	
 
\section{GPLv2~\S6: GPL, My One and Only}
 
\label{GPLv2s6}
 

	
 
A point that was glossed over in Section~\ref{GPLv2s4}'s discussion of GPLv2~\S4
 
was the irrevocable nature of the GPL\@. The GPLv2 is indeed irrevocable,
 
and it is made so formally by GPLv2~\S6.
 

	
 
The first sentence in GPLv2~\S6 ensures that as software propagates down the
 
distribution chain, that each licensor can pass along the license to each
 
new licensee.  Under GPLv2~\S6, the act of distributing automatically grants a
 
license from the original licensor to the next recipient.  This creates a
 
chain of grants that ensure that everyone in the distribution has rights
 
under the GPLv2\@.  In a mathematical sense, this bounds the bottom ---
 
making sure that future licensees get no fewer rights than the licensee before.
 

	
 
The second sentence of GPLv2~\S6 does the opposite; it bounds from the top.  It
 
prohibits any licensor along the distribution chain from placing
 
additional restrictions on the user.  In other words, no additional
 
requirements may trump the rights and freedoms given by GPLv2\@.
 

	
 
The final sentence of GPLv2~\S6 makes it abundantly clear that no individual
 
entity in the distribution chain is responsible for the compliance of any
 
other.  This is particularly important for noncommercial users who have
 
passed along a source offer under GPLv2~\S3(c), as they cannot be assured that
 
the issuer of the offer will honor their GPLv2~\S3 obligations.
 

	
 
In short, GPLv2~\S6 says that your license for the software is your one and
 
only copyright license allowing you to copy, modify and distribute the
 
software.
 

	
 
\section{GPLv2~\S7: ``Give Software Liberty or Give It Death!''}
 
\label{GPLv2s7}
 

	
 
In essence, GPLv2~\S7 is a verbosely worded way of saying for non-copyright
 
systems what GPLv2~\S6 says for copyright.  If there exists any reason that a
 
distributor knows of that would prohibit later licensees from exercising
 
their full rights under GPL, then distribution is prohibited.
 

	
 
Originally, this was designed as the title of this section suggests --- as
 
a last ditch effort to make sure that freedom was upheld.  However, in
 
modern times, it has come to give much more.  Now that the body of GPL'd
 
software is so large, patent holders who would want to be distributors of
 
GPL'd software have a tough choice.  They must choose between avoiding
 
distribution of GPL'd software that exercises the teachings of their
 
patents, or grant a royalty-free, irrevocable, non-exclusive license to
 
those patents.  Many companies have chosen the latter.
 

	
 
Thus, GPLv2~\S7 rarely gives software death by stopping its distribution.
 
Instead, it is inspiring patent holders to share their patents in the same
 
freedom-defending way that they share their copyrighted works.
 

	
 
\section{GPLv2~\S8: Excluding Problematic Jurisdictions}
 
\label{GPLv2s8}
 

	
 
GPLv2~\S8 is rarely used by copyright holders.  Its intention is that if a
 
particular country, say Unfreedonia, grants particular patents or allows
 
copyrighted interfaces (no country to our knowledge even permits those
 
yet), that the GPLv2'd software can continue in free and unabated
 
distribution in the countries where such controls do not exist.
 

	
 
As far as is currently known, GPLv2~\S8 has very rarely been formally used by
 
copyright holders.  Admittedly, some have used GPLv2~\S8 to explain various
 
odd special topics of distribution (usually related in some way to
 
GPLv2~\S7).  However, generally speaking, this section is not proven
 
particularly useful in the more than two decades of GPLv2 history.
 

	
 
Meanwhile, despite many calls by the FSF (and others) for those licensors who
 
explicitly use this section to come forward and explain their reasoning, no
 
one ever did.  Furthermore, research conducted during the GPLv3 drafting
 
process found exactly one licensor who had invoked this section to add an
 
explicit geographical distribution limitation, and the reasoning for that one
 
invocation was not fitting with FSF's intended spirit of GPLv2~\S8.  As such,
 
GPLv2~\S8 was not included at all in GPLv3.
 

	
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 
\chapter{Odds, Ends, and Absolutely No Warranty}
 

	
 
GPLv2~\S\S0--7 constitute the freedom-defending terms of the GPLv2.  The remainder
 
of the GPLv2 handles administrive and issues concerning warranties and
 
of the GPLv2 handles administrivia and issues concerning warranties and
 
liability.
 

	
 
\section{GPLv2~\S9: FSF as Stewards of GPL}
 
\label{GPLv2s9}
 

	
 
FSF reserves the exclusive right to publish future versions of the GPL\@;
 
GPLv2~\S9 expresses this.  While the stewardship of the copyrights on the body
 
of GPL'd software around the world is shared among thousands of
 
individuals and organizations, the license itself needs a single steward.
 
Forking of the code is often regrettable but basically innocuous.  Forking
 
of licensing is disastrous.
 

	
 
(Chapter~\ref{tale-of-two-copylefts} discusses more about the various
 
versions of GPL.)
 

	
 
\section{GPLv2~\S10: Relicensing Permitted}
 
\label{GPLv2s10}
 

	
 
GPLv2~\S10 reminds the licensee of what is already implied by the nature of
 
copyright law.  Namely, the copyright holder of a particular software
 
program has the prerogative to grant alternative agreements under separate
 
copyright licenses.
 

	
 
\section{GPLv2~\S11: No Warranty}
 
\label{GPLv2s11}
 

	
 
Most warranty disclaimer language shout at you.  The
 
\href{http://www.law.cornell.edu/ucc/2/2-316}{Uniform Commercial
 
  Code~\S2-316} requires that disclaimers of warranty be ``conspicuous''.
 
There is apparently general acceptance that \textsc{all caps} is the
 
preferred way to make something conspicuous, and that has over decades worked
 
its way into the voodoo tradition of warranty disclaimer writing.
 

	
 
That said, there is admittedly some authority under USA law suggesting that
 
effective warranty disclaimers that conspicuousness can be established by
 
capitalization and is absent when a disclaimer has the same typeface as the
 
terms surrounding it (see \textit{Stevenson v.~TRW, Inc.}, 987 F.2d 288, 296
 
(5th Cir.~1993)).  While GPLv3's drafters doubted that such authority would
 
apply to copyright licenses like the GPL, the FSF has nevertheless left
 
warranty and related disclaimers in \textsc{all caps} throughout all versions
 
of GPL\@\footnote{One of the authors of this tutorial, Bradley M.~Kuhn, has
 
  often suggested the aesthetically preferable compromise of a
 
  \textsc{specifically designed ``small caps'' font, such as this one, as an
 
    alternative to} WRITING IN ALL CAPS IN THE DEFAULT FONT (LIKE THIS),
 
  since the latter adds more ugliness than conspicuousness.  Kuhn once
 
  engaged in reversion war with a lawyer who disagreed, but that lawyer never
 
  answered Kuhn's requests for case law that argues THIS IS INHERENTLY MORE
 
  CONSPICUOUS \textsc{Than this is}.}.
 

	
 
Some have argued the GPL is unenforceable in some jurisdictions because
 
its disclaimer of warranties is impermissibly broad.  However, GPLv2~\S11
 
contains a jurisdictional savings provision, which states that it is to be
 
interpreted only as broadly as allowed by applicable law.  Such a
 
provision ensures that both it, and the entire GPL, is enforceable in any
 
jurisdiction, regardless of any particular law regarding the
 
permissibility of certain warranty disclaimers.
 

	
 
Finally, one important point to remember when reading GPLv2~\S11 is that GPLv2~\S1
 
permits the sale of warranty as an additional service, which GPLv2~\S11 affirms.
 

	
 
\section{GPLv2~\S12: Limitation of Liability}
 
\label{GPLv2s12}
 

	
 
There are many types of warranties, and in some jurisdictions some of them
 
cannot be disclaimed.  Therefore, usually agreements will have both a
 
warranty disclaimer and a limitation of liability, as we have in GPLv2~\S12.
 
GPLv2~\S11 thus gets rid of all implied warranties that can legally be
 
disavowed. GPLv2~\S12, in turn, limits the liability of the actor for any
 
warranties that cannot legally be disclaimed in a particular jurisdiction.
 

	
 
Again, some have argued the GPL is unenforceable in some jurisdictions
 
because its limitation of liability is impermissibly broad. However, \S
 
12, just like its sister, GPLv2~\S11, contains a jurisdictional savings
 
provision, which states that it is to be interpreted only as broadly as
 
allowed by applicable law.  As stated above, such a provision ensures that
 
both GPLv2~\S12, and the entire GPL, is enforceable in any jurisdiction,
 
regardless of any particular law regarding the permissibility of limiting
 
liability.
 

	
 
So end the terms and conditions of the GNU General Public License.
 

	
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 
\chapter{GPL Version 3}
 
\label{GPLv3}
 

	
 
This chapter discusses the text of GPLv3.  Much of this material herein
 
includes text that was adapted (with permission) from text that FSF
 
originally published as part of the so-called ``rationale documents'' for the
 
various discussion drafts of GPLv3.
 

	
 
The FSF ran a somewhat public process to develop GPLv3, and it was the first
 
attempt of its kind to develop a Free Software license this way.  Ultimately,
 
RMS was the primary author of GPLv3, but he listened to feedback from all
 
sorts of individuals and even for-profit companies.  Nevertheless, in
 
attempting to understand GPLv3 after the fact, the materials available from
 
the GPLv3 process have a somewhat ``drinking from the firehose'' effect.
 
This chapter seeks to explain GPLv3 to newcomers, who perhaps are familiar
 
with GPLv2 and who did not participate in the GPLv3 process.
 

	
 
Those who wish to drink from the firehose and take a diachronic approach to
 
GPLv3 study by reading the step-by-step public drafting process of the GPLv3 (which
 
occurred from Monday 16 January 2006 through Monday 19 November 2007) should
 
visit \url{http://gplv3.fsf.org/}.
 

	
 
\section{Understanding GPLv3 As An Upgraded GPLv2}
 

	
 
Ultimately, GPLv2 and GPLv3 co-exist as active licenses in regular use.  As
 
discussed in Chapter~\ref{tale-of-two-copylefts}, GPLv1 was never regularly
 
used alongside GPLv2.  However, given GPLv2's widespread popularity and
 
existing longevity by the time GPLv3 was published, it is not surprising that
 
some licensors still prefer GPLv2-only or GPLv2-or-later.  GPLv3 gained major
 
adoption by many projects, old and new, but many projects have not upgraded
 
due to (in some cases) mere laziness and (in other cases) policy preference
 
for some of GPLv2's terms and/or policy opposition to GPLv3's terms.
 

	
 
Given this ``two GPLs world'' is reality, it makes sense to consider GPLv3 in
 
terms of how it differs from GPLv2.  Also, most of the best GPL experts in
 
the world must deal regularly with both licenses, and admittedly have decades
 
of experience with GPLv2 while the most experience with GPLv3 that's possible
 
is by default less than a decade.  These two factors usually cause even new
 
students of GPL to start with GPLv2 and move on to GPLv3, and this tutorial
 
follows that pattern.
 

	
 
Overall, the changes made in GPLv3 admittedly \textit{increased} the
 
complexity of the license.  The FSF stated at the start of the GPLv3 process
 
that they would have liked to oblige those who have asked for a simpler and
 
shorter GPL\@.  Ultimately, the FSF gave priority to making GPLv3 a better
 
copyleft license in the spirit of past GPL's.  Obsession for concision should
 
never trump software freedom.
 

	
 
The FSF had many different, important goals in seeking to upgrade to GPLv3.
 
However, one important goal that is often lost in the discussion of policy
 
minutia is a rather simple but important issue.  Namely, FSF sought to assure
 
that GPLv3 was more easily internationalized than GPLv2.  In particular, the
 
FSF sought to ease interpretation of GPL in other countries by replacement of
 
USA-centric\footnote{See Section~\ref{non-usa-copyright} of this tutorial for
 
  a brief discussion about non-USA copyright systems.}  copyright phrases and
 
wording with neutral terminology rooted in description of behavior rather
 
than specific statue.  As can be seen in the section-by-section discussion of
 
GPLv3 that follows, nearly every section had changes related to issues of
 
internationalization.
 
 
 
\section{GPLv3~\S0: Giving In On ``Defined Terms''}
 
\label{GPLv3s0}
 

	
 
One of lawyers' most common complaints about GPLv2 is that defined terms in
 
the document appear throughout.  Most licenses define terms up-front.
 
However, the GPL was always designed both as a document that should be easily
 
understood both by lawyers and by software developers: it is a document
 
designed to give freedom to software developers and users, and therefore it
 
should be comprehensible to that constituency.
 

	
 
Interestingly enough, one coauthor of this tutorial who is both a lawyer and
 
a developer pointed out that in law school, she understood defined terms more
 
quickly than other law students precisely because of her programming
 
background.  For developers, having \verb0#define0 (in the C programming
 
language) or other types of constants and/or macros that automatically expand
 
in the place where they are used is second nature.  As such, adding a defined
 
terms section was not terribly problematic for developers, and thus GPLv3
 
adds one.  Most of these defined terms are somewhat straightforward and bring
 
forward better worded definitions from GPLv2.  Herein, this tutorial
 
discusses a few of the new ones.
 

	
 
GPLv3~\S0 includes definitions of five new terms not found in any form in
 
GPLv2: ``modify'' ``covered work'', ``propagate'', ``convey'', and
 
``Appropriate Legal Notices''. 
 

	
 
\subsection{Modify and the Work Based on the Program}
 

	
 
GPLv2 included a defined term, ``work based on the Program'', but also used
 
the term ``modify'' and ``based on'' throughout the license.  GPLv2's ``work
 
based on the Program'' definition made use of a legal term of art,
 
``derivative work'', which is peculiar to USA copyright law.  However,
 
ironically, most criticism of USA-specific legal terminology in GPLv2's
 
``work based on the Program'' definition historically came not primarily from
 
readers outside the USA, but from those within it\footnote{The FSF noted in
 
  that it did not generally agree with these views, and expressed puzzlement
 
  by the energy with which they were expressed, given the existence of many
 
  other, more difficult legal issues implicated by the GPL.  Nevertheless,
 
  the FSF argued that it made sense to eliminate usage of local copyright
 
  terminology to good effect.}.  Admittedly, even though differently-labeled
 
concepts corresponding to the derivative work are recognized in all copyright
 
law systems, these counterpart concepts might differ to some degree in scope
 
and breadth from the USA derivative work.
 

	
 
The goal and intention of GPLv2 was always to cover all rights governed by
 
relevant copyright law, in the USA and elsewhere.  GPLv3 therefore takes the
 
task of internationalizing the license further by removing references to
 
derivative works and by providing a more globally useful definition.  The new
 
definition returns to the common elements of copyright law.  Copyright
 
holders of works of software have the exclusive right to form new works by
 
modification of the original --- a right that may be expressed in various
 
ways in different legal systems.  GPLv3 operates to grant this right to
 
successive generations of users (particularly through the copyleft conditions
 
set forth in GPLv3~\S5, as described later in this tutorial in its
 
\S~\ref{GPLv3s5}).  Here in GPLv3~\S0, ``modify'' refers to basic copyright
 
rights, and then this definition of ``modify'' is used to define ``modified
 
version of'' and ``work based on'' as synonyms.
 

	
 
\subsection{The Covered Work}
 

	
 
GPLv3 uses a common license drafting technique of building upon simpler
 
definitions to make complex ones.  The Program is a defined term found
 
throughout GPLv2, and the word ``covered'' and the phrase ``covered by this
 
license'' are used in tandem with the Program in GPLv2, but not as part of a
 
definition.  GPLv3 offers a single term ``covered work'', which enables some
 
of the wording in GPLv3 to be simpler and clearer than its GPLv2
 
counterparts.
 

	
 
\subsection{Propagate}
 

	
 
The term ``propagate'' serves two purposes.  First, ``propagate'' provides a
 
simple and convenient means for distinguishing between the kinds of uses of a
 
work that GPL imposes conditions on and the kinds of uses that GPL does not
 
(for the most part) impose conditions on.
 

	
 
Second, ``propagate'' helps globalize GPL in its wording and effect.  When a
 
work is GPL'd, the copyright law of some particular country will govern
 
certain legal issues arising under the license.  A term like ``distribute''
 
(or its equivalent in languages other than English) is used in several
 
national copyright statutes.  Yet, practical experience with GPLv2 revealed
 
the awkwardness of using the term ``distribution'' in a license intended for
 
global use: the scope of ``distribution'' in the copyright context can differ
 
from country to country.  The GPL never necessarily intended the specific
 
meaning of ``distribution'' that exists under USA (or any other country's)
 
copyright law.
 

	
 
Indeed, even within a single country and language, the term distribution may
 
be ambiguous; as a legal term of art, distribution varies significantly in
 
meaning among those countries that recognize it.  For example, comments
 
during GPLv3's drafting process indicated that in at least one country,
 
distribution may not include network transfers of software but may include
 
interdepartmental transfers of physical copies within an organization.
 
Meanwhile, the copyright laws of many countries, as well as certain
 
international copyright treaties, recognize ``making available to the
 
public'' or ``communication to the public'' as one of the exclusive rights of
 
copyright holders.
 

	
 
Therefore, the GPL defines the term ``propagate'' by reference to activities
 
that require permission under ``applicable copyright law'', but excludes
 
execution and private modification from the definition.  GPLv3's definition
 
also gives examples of activities that may be included within ``propagation''
 
but it also makes clear that, under the copyright laws of a given country,
 
``propagation'' may include other activities as well.
 

	
 
Thus, propagation is defined by behavior, and not by categories drawn from
 
some particular national copyright statute.  This helps not only with
 
internationalization, but also factually-based terminology aids in
 
developers' and users' understanding of the GPL\@.
 

	
 
\subsection{Convey}
 

	
 
Further to this point, a subset of propagate --- ``convey'' --- is defined.
 
Conveying includes activities that constitute propagation of copies to
 
others.  As with the definition of propagate, GPLv3 thus addresses transfers
 
of copies of software in behavioral rather than statutory terms.  
 

	
 
\subsection{Appropriate Legal Notices}
 

	
 
GPLv2 used the term ``appropriate copyright notice and disclaimer of
 
warranty'' in two places, which is a rather bulk term.  Also, experience with
 
GPLv2 and other licenses that grant software freedom showed throughout the
 
1990s that the scope of types of notices that need preservation upon
 
conveyance were more broad that merely the copyright notices.  The
 
Appropriate Legal Notice definition consolidates the material that GPLv2
 
traditionally required preserved into one definition.
 

	
 
\subsection{Other Defined Terms}
 

	
 
Note finally that not all defined terms in GPLv3 appear in GPLv3~\S0.
 
Specifically, those defined terms that are confined in use to a single
 
section are defined in the section in which they are used, and GPLv3~\S1
 
contains those definitions focused on source code.  In this tutorial, those
 
defined terms are discussed in the section where they are defined and/or
 
used.
 

	
 
\section{GPLv3~\S1: Understanding CCS}
 
\label{GPLv3s1}
 

	
 
Ensuring that users have the source code to the software they receive and the
 
freedom to modify remains the paramount right embodied in the Free Software
 
Definition (found in \S~\ref{Free Software Definition} of this tutorial).  As
 
such, GPLv3~\S1 is likely one of the most important sections of GPLv3, as it
 
contains all the defined terms related to this important software freedom.
 

	
 
\subsection{Source Code Definition}
 

	
 
First, GPLv3~\S1 retains GPLv2's definition of ``source code'' and adds an
 
explicit definition of ``object code'' as ``any non-source version of a
 
work''.  Object code is not restricted to a narrow technical meaning and is
 
understood broadly to include any form of the work other than the preferred
 
form for making modifications to it.  Object code therefore includes any kind
 
of transformed version of source code, such as bytecode or minified
 
Javascript.  The definition of object code also ensures that licensees cannot
 
escape their obligations under the GPL by resorting to shrouded source or
 
obfuscated programming.
 

	
 
\subsection{CCS Definition}
 

	
 
The definition of CCS\footnote{Note that the preferred term for those who
 
  work regularly with both GPLv2 and GPLv3 is ``Complete Corresponding
 
  Source'', abbreviated to ``CCS''.  Admittedly, the word ``complete'' no
 
  longer appears in GPLv3 (which uses the word ``all'' instead).  However,
 
  both GPLv2 and the early drafts of GPLv3 itself used the word ``complete'',
 
  and early GPLv3 drafts even called this defined term ``Complete
 
  Corresponding Source''.  Meanwhile, use of the acronym ``CCS'' (sometimes,
 
  ``C\&CS'') was so widespread among GPL enforcers that its use continues
 
  even though GPLv3-focused experts tend to say just the defined term of
 
  ``Corresponding Source''.}, or, as GPLv3 officially calls it,
 
``Corresponding Source'' in GPLv3~\S1\P4 is possibly the most complex
 
definition in the license.
 

	
 
The CCS definition is broad so as to protect users' exercise of their rights
 
under the GPL\@.  The definition includes with particular examples to remove
 
any doubt that they are to be considered CCS\@.  GPLv3 seeks to make it
 
completely clear that a licensee cannot avoid complying with the requirements
 
of the GPL by dynamically linking a subprogram component to the original
 
version of a program.  The example also clarifies that the shared libraries
 
and dynamically linked subprograms that are included in Corresponding Source
 
are those that the work is ``specifically'' designed to require, which
 
clarifies that they do not include libraries invoked by the work that can be
 
readily substituted by other existing implementations.  While copyleft
 
advocates never doubted this was required under GPLv2's definition of CCS,
 
GPLv3 makes it abundantly clear with an extra example.
 

	
 
The GPL, as always, seeks to ensure users are truly in a position to install and
 
run their modified versions of the program; the CCS definition is designed to
 
be expansive to ensure this software freedom.  However, although the
 
definition of CCS is expansive, it is not sufficient to protect users'
 
freedoms in many circumstances.  For example, a GPL'd program, or a modified
 
version of such a program, might be locked-down and restricted.  The
 
requirements in GPLv3~\S6 (discussed in Section~\ref{GPLv3s6} of this
 
tutorial) handle that issue.  (Early drafts of GPLv3 included those
 
requirements in the definition of CCS; however, given that the lock-down
 
issue only comes up in distribution of object code, it is more logical to
 
place those requirements with the parts of GPLv3 dealing directly with object
 
code distribution).
 

	
 
The penultimate paragraph in GPLv3\S2 notes that GPLv3's CCS definition does
 
not require source that can be automatically generated.  Many code
 
generators, preprocessors and take source code as input and sometimes even
 
have output that is still source code.  Source code should always be whatever
 
the original programmer preferred to modify.
 

	
 
GPLv3\S1's final paragraph removes any ambiguity about what should be done on
 
source-only distributions.  Specifically, the right to convey source code
 
that does not compile, does not work, or otherwise is experimental
 
in-progress work is fully permitted, \textit{provided that} no object code
 
form is conveyed as well.  Indeed, when combined with the permissions in
 
GPLv3\S~5, it is clear that if one conveys \textit{only} source code, one can
 
never be required to provide more than that.  One always has the right to
 
modify a source code work by deleting any part of it, and there can be no
 
requirement that free software source code be a whole functioning program.
 

	
 
\subsection{The System Library Exception}
 

	
 
The previous section skipped over one part of the CCS definition, the
 
so-called system library exception.  The ``System Libraries'' definition (and
 
the ``Standard Interface'' and ``Major Component'' definitions, which it
 
includes) are designed as part
 
to permit certain distribution arrangements that are considered reasonable by
 
copyleft advocates.  The system library exception is designed to allow
 
copylefted software to link with these libraries when such linking would hurt
 
software freedom more than it would hurt proprietary software.
 

	
 
The system library exception has two parts.  Part (a) rewords the GPLv2
 
exception for clarity replacing GPLv2's words ``unless that component itself
 
accompanies the executable'' with ``which is not part of the Major
 
Component''.  The goal here is to not require disclosure of source code of
 
certain libraries, such as necessary Microsoft Windows DLLs (which aren't
 
part of Windows' kernel but accompany it) that are required for functioning
 
of copylefted programs compiled for Windows.
 

	
 
However, in isolation, (a) would be too permissive, as it would sometimes
 
allowing distributors to evade important GPL requirements.  Part (b) reigns
 
in (a).  Specifically, (b) specifies only a few functionalities that a
 
system library may provide and still qualify for the exception.  The goal is
 
to ensure system libraries are truly adjunct to a major essential operating
 
system component, compiler, or interpreter.  The more low-level the
 
functionality provided by the library, the more likely it is to be qualified
 
for this exception.
 

	
 
Admittedly, the system library exception is a frequently discussed topic of
 
obsessed GPL theorists.  The amount that has been written on the system
 
library exception (both the GPLv2 and GPLv3 versions of it), if included
 
herein,  could easily increase this section of the tutorial to a length
 
greater than all the others.
 

	
 
Like any exception to the copyleft requirements of GPL, would-be GPL
 
violators frequently look to the system library exception as a potential
 
software freedom circumvention technique.  When considering whether or not a
 
library qualifies for the system library exception, here is a pragmatic
 
thesis to consider, based on the combined decades of experience in GPL
 
interpretation of this tutorial's authors: the harder and more strained the
 
reader must study and read the system library exception, the more likely it
 
is that the library in question does not qualify for it.
 

	
 
\section{GPLv3~\S2: Basic Permissions}
 

	
 
GPLv3~\S2 can roughly be considered as an equivalent to GPLv2~\S0 (discussed
 
in \S~\ref{GPLv2s0} of this tutorial).  However, the usual style of
 
improvements found in GPLv3 are found here as well.  For example, the first
 
sentence of GPLv3~\S2 furthers the goal internationalization.  Under the
 
copyright laws of some countries, it may be necessary for a copyright license
 
to include an explicit provision setting forth the duration of the rights
 
being granted. In other countries, including the USA, such a provision is
 
unnecessary but permissible.
 

	
 
GPLv3~\S2\P1 also acknowledges that licensees under the GPL enjoy rights of
 
copyright fair use, or the equivalent under applicable law.  These rights are
 
compatible with, and not in conflict with, the freedoms that the GPL seeks to
 
protect, and the GPL cannot and should not restrict them.
 

	
 
However, note that (sadly to some copyleft advocates) the unlimited freedom
 
to run is confined to the \textit{unmodified} Program.  This confinement is
 
unfortunately necessary since Programs that do not qualify as a User Product
 
in GPLv3~\S6 (see \S~\ref{user-product} in this tutorial) might have certain
 
unfortunate restrictions on the freedom to run.\footnote{See
 
  \S~\ref{freedom-to-run} of this tutorial for the details on ``the freedom to
 
  run''.}
 

	
 
GPLv3~\S2\P2 distinguishes between activities of a licensee that are
 
permitted without limitation and activities that trigger additional
 
requirements.  Specifically, GPLv3~\S2\P2 guarantees the basic freedoms of
 
privately modifying and running the program.
 

	
 
Also, GPLv3~\S2\P2 gives an explicit permission for a client to provide a
 
copy of its modified software to a contractor exclusively for that contractor
 
to modify it further, or run it, on behalf of the client.  However, the
 
client can \textit{only} exercise this control over its own copyrighted
 
changes to the GPL-covered program.  The parts of the program it obtained
 
from other contributors must be provided to the contractor with the usual GPL
 
freedoms.  Thus, GPLv3 permits users to convey covered works to contractors
 
operating exclusively on the users' behalf, under the users' direction and
 
control, and to require the contractors to keep the users' copyrighted
 
changes confidential, but \textit{only if} the contractor is limited to acting
 
on the users' behalf (just as the users' employees would have to act).
 

	
 
The strict conditions in this ``contractors provision'' are needed so that it
 
cannot be twisted to fit other activities, such as making a program available
 
to downstream users or customers.  By making the limits on this provision
 
very narrow, GPLv3 ensures that, in all other cases, contractor gets the
 
full freedoms of the GPL that they deserve.
 

	
 
The FSF was specifically asked to add this ``contractors provisions'' by
 
large enterprise users of Free Software, who often contract with non-employee
 
developers, working offsite, to make modifications intended for the user's
 
private or internal use, and often arrange with other companies to operate
 
their data centers.  Whether GPLv2 permits these activities is not clear and
 
may depend on variations in copyright law in different jurisdictions.  The
 
practices seem basically harmless, so FSF decided to make it clear they are
 
permitted.
 

	
 
GPLv3~\S2's final paragraph includes an explicit prohibition of sublicensing.
 
This provision ensures that GPL enforcement is always by the copyright
 
holder.  Usually, sublicensing is regarded as a practical convenience or
 
necessity for the licensee, to avoid having to negotiate a license with each
 
licensor in a chain of distribution.  The GPL solves this problem in another
 
way --- through its automatic licensing provision found in GPLv3~\S10 (which
 
is discussed in more detail in \S~\ref{GPLv3s10} of this tutorial).
 

	
 
\section{GPLv3's views on DRM and Device Lock-Down}
 
\label{GPLv3-drm}
 

	
 
The issues of DRM, device lock-down and encryption key disclosure were the
 
most hotly debated during the GPLv3 process.  FSF's views on this were sadly
 
frequently misunderstood and, comparing the provisions related to these
 
issues in the earliest drafts of GPLv3 to  the final version of GPLv3 shows
 
the FSF's willingness to compromise on tactical issues to reach the larger
 
goal of software freedom.
 

	
 
Specifically, GPLv3 introduced provisions that respond to the growing
 
practice of distributing GPL-covered programs in devices that employ
 
technical means to restrict users from installing and running modified
 
versions.  This practice thwarts the expectations of developers and users
 
alike, because the right to modify is one of the core freedoms the GPL is
 
designed to secure.
 

	
 
Technological measures to defeat users' rights.  These measures are often
 
described by such Orwellian phrases, such as ``digital rights management,''
 
which actually means limitation or outright destruction of users' legal
 
rights, or ``trusted computing,'' which actually means selling people
 
computers they cannot trust.  However, these measures are alike in one basic
 
respect.  They all employ technical means to turn the system of copyright law
 
(where the powers of the copyright holder are limited exceptions to general
 
freedom) into a virtual prison, where everything not specifically permitted
 
is utterly forbidden.  This system of ``para-copyright'' was created well
 
after GPLv2 was written --- initially through legislation in the USA and the
 
EU, and later in other jurisdictions as well.  This legislation creates
 
serious civil or even criminal penalties to escape from these restrictions
 
(commonly and aptly called ``jail-breaking a device''), even where the
 
purpose in doing so is to restore the users' legal rights that the technology
 
wrongfully prevents them from exercising.
 

	
 
GPLv2 did not address the use of technical measures to take back the rights
 
that the GPL granted, because such measures did not exist in 1991, and would
 
have been irrelevant to the forms in which software was then delivered to
 
users.  GPLv3 addresses these issues, particularly because copylefted
 
software is ever more widely embedded in devices that impose technical
 
limitations on the user's freedom to change it.
 

	
 
However, FSF always made a clear distinction to avoid conflating these
 
``lock-down'' measures with legitimate applications that give users control,
 
as by enabling them to choose higher levels of system or data security within
 
their networks, or by allowing them to protect the security of their
 
communications using keys they can generate or copy to other devices for
 
sending or receiving messages.  Such technologies present no obstacles to
 
software freedom and the goals of copyleft.
 

	
 
The public GPLv3 drafting process sought to balance these positions of
 
copyleft advocates with various desperate views of the larger
 
Free-Software-using community.  Ultimately, FSF compromised to the GPLv3\S3
 
and GPLv3\S6 provisions that, taken together, are a minimalist set of terms
 
sufficient to protect the software freedom against the threat of invasive
 
para-copyright.
 

	
 
The compromises made were ultimately quite reasonable.  The primary one is
 
embodied in GPLv3\S6's ``User Product'' definition (see \S~\ref{user-product}
 
in this tutorial for details).  Additionally, some readers of early GPLv3
 
drafts seem to have assumed GPLv3 contained a blanket prohibition on DRM; but
 
it does not.  In fact, no part of GPLv3 forbids DRM regarding non-GPL'd
 
works; rather, GPLv3 forbids the use of DRM specifically to lock-down
 
restrictions on users' ability to install modified versions of the GPL'd
 
software itself, but again, \textit{only} with regard to User Products.
 

	
 
\section{GPLv3~\S3: What Hath DMCA Wrought}
 
\label{GPLv3s3}
 

	
 
As discussed in \S~\ref{software-and-non-copyright} of this tutorial,
 
\href{http://www.law.cornell.edu/uscode/text/17/1201}{17 USC~\S1201} and
 
relate sections\footnote{These sections of the USC are often referred to as
 
  the ``Digital Millennium Copyright Act'', or ``DMCA'', as that was the name
 
  of the bill that so-modified these sections of the USC\@.} prohibits users
 
from circumventing technological measures that implement DRM\@.  Since this
 
is part of copyright law and the GPL is primarily a copyright license, and
 
since what the DMCA calls ``circumvention'' is simply ``modifying the
 
software'' under the GPL, GPLv3 must disclaim that such anti-circumvention
 
provisions are not applicable to the GPLv3'd software.  GPLv3\S3 shields
 
users from being subjected to liability under anti-circumvention law for
 
exercising their rights under the GPL, so far as the GPL can do so.
 

	
 
First, GPLv3\S3\P1 declares that no GPL'd program is part of an effective
 
technological protection measure, regardless of what the program does.  Early
 
drafts of GPLv3\S3\P1 referred directly to the DMCA, but the final version
 
instead includes instead an international legal reference to
 
anticircumvention laws enacted pursuant to the 1996 WIPO treaty and any
 
similar laws.  Lawyers outside the USA worried that a USA statutory reference
 
could be read as indicating a choice for application of USA law to the
 
license as a whole.  While the FSF did not necessarily agree with that view,
 
the FSF decided anyway to refer to the WIPO treaty rather than DMCA, since
 
several national anticircumvention laws were (or will likely be) structured
 
more similarly to the anticircumvention provisions of the DMCA in their
 
implementation of WIPO\@.  Furthermore, the addition of ``or similar laws''
 
provides an appropriate catch-all.
 

	
 
Furthermore, GPLv3\S3\P2 states precisely that a conveying party waives the
 
power to forbid circumvention of technological measures only to the extent
 
that such circumvention is accomplished through the exercise of GPL rights in
 
the conveyed work.  GPLv3\S3\P2 makes clear that the referenced ``legal
 
rights'' are specifically rights arising under anticircumvention law.  and
 
refers to both the conveying party's rights and to third party rights, as in
 
some cases the conveying party will also be the party legally empowered to
 
enforce or invoke rights arising under anticircumvention law.
 

	
 
These disclaimers by each licensor of any intention to use GPL'd software to
 
stringently control access to other copyrighted works should effectively
 
prevent any private or public parties from invoking DMCA-like laws against
 
users who escape technical restriction measures implemented by GPL'd
 
software.
 

	
 
\section{GPLv3~\S4: Verbatim Copying}
 
\label{GPLv3s4}
 

	
 
GPLv3~\S4 is a revision of GPLv2~\S1 (as discussed in \S~\ref{GPLv2s1} of
 
this tutorial).   There are almost no changes to this section from the
 
GPLv2~\S1, other than to use the new defined terms.
 

	
 
The only notable change of ``a fee'' to ``any price or no price'' in the
 
first sentence of GPLv3\S4\P2.  The GPLv2\S1\P1 means that the GPL permits
 
one to charge money for the distribution of software.  Despite efforts by
 
copyleft advocates to explain this in GPLv2 itself and in other documents,
 
there are evidently some people who still believe that GPLv2 allows charging
 
for services but not for selling copies of software and/or that the GPL
 
requires downloads to be gratis.  Perhaps this is because GPLv2 referred to
 
charging a ``fee''; the term ``fee'' is generally used in connection with
 
services.
 

	
 
GPLv2's wording also referred to ``the physical act of transferring.''  The
 
intention was to distinguish charging for transfers from attempts to impose
 
licensing fees on all third parties.  ``Physical'' might be read, however, as
 
suggesting ``distribution in a physical medium only''.
 

	
 
To address these two issues, GPLv3 says ``price'' in place of ``fee,'' and
 
removes the term ``physical.''
 

	
 
GPLv3~\S4 has also been revised from its corresponding section in GPLv2 in
 
light of the GPLv3~\S7 (see \S~\ref{GPLv3s7} in this tutorial for more).
 
Specifically, a distributor of verbatim copies of the program's source code
 
must obey any existing additional terms that apply to parts of the program
 
pursuant to GPLv3~\S7.  In addition, the distributor is required to keep
 
intact all license notices, including notices of such additional terms.
 

	
 
Finally, there is no harm in explicitly pointing out what ought to be
 
obvious: that those who convey GPL-covered software may offer commercial
 
services for the support of that software.
 

	
 
\section{GPLv3~\S5: Modified Source}
 
\label{GPLv3s5}
 

	
 
GPLv3\S5 is the rewrite of GPLv2\S2, which was discussed in \S~\ref{GPLv2s2}
 
of this tutorial.  This section discusses the changes found in GPLv3\S5
 
compared to GPLv2\S2.
 

	
 
GPLv3\S5(a) still requires modified versions be marked with ``relevant
 
date'', but no longer says ``the date of any change''.  The best practice is
 
to include the date of the latest and/or most significant changes and who
 
made those.  Of course, compared to its GPLv2\S2(a), GPLv3\S5(a) slightly
 
relaxes the requirements regarding notice of changes to the program.  In
 
particular, the modified files themselves need no longer be marked.  This
 
reduces administrative burdens for developers of modified versions of GPL'd
 
software.
 

	
 
GPLv3\S5(b) is a new but simple provision. GPLv3\S5(b)  requires that the
 
license text itself must be unmodified (except as permitted by GPLv3\S7; see
 
\S~\ref{GPLv3s7} in this tutorial).  Furthermore, it  removes any perceived
 
conflict between the words ``keep intact all notices'' in GPLv3\S4, since
 
operating under GPLv3\S5 still includes all the requirements of GPLv3\S4 by
 
reference.
 

	
 
GPLv3\S5(c) is the primary source-code-related copyleft provision of GPL. (The
 
object-code-related copyleft provisions are in GPLv3\S6, discussed in
 
\S~\ref{GPLv3s6} of this tutorial).  Compared to GPLv2\S2(b), GPLv3\S5(c)
 
states that the GPL applies to the whole of the work.  Such was stated
 
already in GPLv2\S2(b), in ``in whole or in part'', but this simplified
 
wording makes it clear the entire covered work
 

	
 
Another change in GPLv3\S5(c) is the removal of the
 
words ``at no charge,'' which was often is misunderstood upon na\"{i}ve
 
reading of in GPLv2\S(b) (as discussed in \S~\ref{GPLv2s2-at-no-charge} of this
 
tutorial).
 

	
 
%  FIXME-LATER: Write up something on 5d, and related it to Appropriate Legal Notices.
 

	
 

	
 
Note that of GPLv2~\S2's penultimate and ante-penultimate paragraphs are now
 
handled adequately by the definitions in GPLv3\S0 and as such, have no direct
 
analogs in GPLv3.
 

	
 
GPLv2~\S2's final paragraph, however, is reworded and expanded into the final
 
paragraph of GPLv3\S5, which now also covers issues related to copyright
 
compilations (but not compilations into object code --- that's in the next
 
section!).  The intent and scope is the same as was intended in GPLv2.
 

	
 
\section{GPLv3~\S6: Non-Source and Corresponding Source}
 
\label{GPLv3s6}
 

	
 
GPLv3~\S6 clarifies and revises GPLv2~\S3.  It requires distributors of GPL'd
 
object code to provide access to the corresponding source code, in one of
 
four specified ways.  As noted in \S~\ref{GPLv3s0}, ``object code'' in GPLv3
 
is defined broadly to mean any non-source version of a work.
 

	
 
% FIXME:  probably mostly still right, needs some updates, though.
 

	
 
GPLv3~\S6(a--b) now apply specifically to distribution of object code in a
 
physical product.  Physical products include embedded systems, as well as
 
physical software distribution media such as CDs.  As in GPLv2~\S3 (discussed
 
in \S~\ref{GPLv2s3} of this tutorial), the distribution of object code may
 
either be accompanied by the machine-readable source code, or it may be
 
accompanied by a valid written offer to provide the machine-readable source
 
code.  However, unlike in GPLv2, that offer cannot be exercised by any third
 
party; rather, only those ``who possesses the object code'' can exercise
 
the offer.  (Note that this is a substantial narrowing of requirements of
 
offer fulfillment, and is a wonderful counterexample to dispute claims that
 
the GPLv3 has more requirements than GPLv2.)
 

	
 
% FIXME:  probably mostly still right, needs some updates, though.
 

	
 
GPLv3~\S6(b) further revises the requirements for the written offer to
 
provide source code. As before, the offer must remain valid for at least
 
three years. In addition, even after three years, a distributor of a product
 
containing GPL'd object code must offer to provide source code for as long as
 
the distributor also continues to offer spare parts or customer support for
 
the product model.  This is a reasonable and appropriate requirement; a
 
distributor should be prepared to provide source code if he or she is
 
prepared to provide support for other aspects of a physical product.
 

	
 
GPLv3~\S6(a--b) clarifies that the medium for software interchange on which
 
the machine-readable source code is provided must be a durable physical
 
medium.  GPLv3~\S6(b)(2), however, permits a distributor to instead offer to
 
provide source code from a network server instead, which is yet another
 
example GPLv3 looser in its requirements than GPLv2 (see
 
\S~\ref{GPLv2s3-medium-customarily} for details).
 

	
 
% FIXME-LATER: more information about source provision, cost of physically
 
% performing, reasonable fees, medium customary clearly being said durable
 
% connecting back to previous text
 

	
 
GPLv3\S6(c) gives narrower permission than GPLv2\S3(c).  The ``pass along''
 
option for GPLv3\S6(c)(1) offers is now available only for individual
 
distribution of object code; moreover, such individual distribution can occur
 
only ``occasionally and noncommercially.''  A distributor cannot comply with
 
the GPL merely by making object code available on a publicly-accessible
 
network server accompanied by a copy of the written offer to provide source
 
code received from an upstream distributor.
 

	
 
%FIXME-LATER: tie back to the discussion of the occasional offer pass along
 
%             stuff in GPLv2 this tutorial.
 

	
 
GPLv3~\S6(d) revises and improves GPLv2~\S3's final paragraph.  When object
 
code is provided by offering access to copy the code from a designated place
 
(such as by enabling electronic access to a network server), the distributor
 
must merely offer equivalent access to copy the source code ``in the same way
 
through the same place''.  This wording also permits a distributor to offer a
 
third party access to both object code and source code on a single network
 
portal or web page, even though the access may include links to different
 
physical servers.  For example, a downstream distributor may provide a link
 
to an upstream distributor's server and arrange with the operator of that
 
server to keep the source code available for copying for as long as the
 
downstream distributor enables access to the object code.  This codifies
 
formally the typical historical interpretation of GPLv2.
 

	
 
% FIXME-LATER: perhaps in enforcement section, but maybe here, note about
 
% ``slow down'' on source downloads being a compliance problem. 
 

	
 
Furthermore, under GPLv3~\S6(d), distributors may charge for the conveyed
 
object code; however, those who pay to obtain the object code must be given
 
equivalent and gratis access to obtain the CCS.  (If distributors convey the
 
object code gratis, distributors must likewise make CCS available without
 
charge.)  Those who do not obtain the object code from that distributors
 
(perhaps because they choose not to pay the fee for object code) are outside
 
the scope of the provision; distributors are under no specific obligation to
 
give CCS to someone who has not purchased an object code download under
 
GPLv3~\S6(d).  (Note: this does not change nor impact any obligations under
 
GPLv3~\S6(b)(2); GPLv3~\S6(d) is a wholly different provision.)
 

	
 
\subsection{GPLv3~\S6(e): Peer-to-Peer Sharing Networks}
 

	
 
Certain decentralized forms of peer-to-peer file sharing present a challenge
 
to the unidirectional view of distribution that is implicit in GPLv2 and
 
Draft 1 of GPLv3.  Identification of an upstream/downstream link in
 
BitTorrent distribution is neither straightforward nor reasonable; such
 
distribution is multidirectional, cooperative and anonymous.  In peer-to-peer
 
distribution systems, participants act both as transmitters and recipients of
 
blocks of a particular file, but they perceive the experience merely as users
 
and receivers, and not as distributors in any conventional sense.  At any
 
given moment of time, most peers will not have the complete file.
 

	
 
Meanwhile, GPLv3~\S6(d) permits distribution of a work in object code form
 
over a network, provided that the distributor offers equivalent access to
 
copy the Corresponding Source Code ``in the same way through the same
 
place''.  This wording might be interpreted to permit peer-to-peer
 
distribution of binaries \textit{if} they are packaged together with the CCS,
 
but such packaging impractical, for at least three reasons.  First, even if
 
the CCS is packaged with the object code, it will only be available to a
 
non-seeding peer at the end of the distribution process, but the peer will
 
already have been providing parts of the binary to others in the network.
 
Second, in practice, peer-to-peer forms of transmission are poorly suited
 
means for distributing CCS.  In large distributions, packaging CCS with the
 
object code may result in a substantial increase in file size and
 
transmission time.  Third, in current practice, CCS packages themselves tend
 
\textit{not} to be transmitted through BitTorrent --- owing to reduced demand
 
-- thus, there generally will be too few participants downloading the same
 
source package at the same time to enable effective seeding and distribution.
 

	
 
GPLv3~\S6(e) addresses this issues.  If a licensee conveys such a work of
 
object code using peer-to-peer transmission, that licensee is in compliance
 
with GPLv3~\S6 if the licensee informs other peers where the object code and
 
its CCS are publicly available at no charge under subsection GPLv3~\S6(d).
 
The CCS therefore need not be provided through the peer-to-peer system that
 
was used for providing the binary.
 

	
 
Second, GPLv3\S9 also clarifies that ancillary propagation of a covered work
 
that occurs as part of the process of peer-to-peer file transmission does not
 
require acceptance, just as mere receipt and execution of the Program does
 
not require acceptance.  Such ancillary propagation is permitted without
 
limitation or further obligation.
 

	
 
% FIXME-LATER: Would be nice to explain much more about interactions between
 
% the various options of GPLv3~\S6(a-e), which might all be in play at once!
 

	
 
\subsection{User Products, Installation Information and Device Lock-Down}
 

	
 
As discussed in \S~\ref{GPLv3-drm} of this tutorial, GPLv3 seeks to thwart
 
technical measures such as signature checks in hardware to prevent
 
modification of GPL'd software on a device.
 

	
 
To address this issue, GPLv3~\S6 requires that parties distributing object
 
code provide recipients with the source code through certain means.  When
 
those distributors pass on the CCS, they are also required to pass on any
 
information or data necessary to install modified software on the particular
 
device that included it.  (This strategy is not unlike that used in LGPLv2.1
 
to enable users to link proprietary programs to modified libraries.)
 

	
 
% FIXME-LATER: LGPLv2.1 section should talk about this explicitly and this
 
%              should be a forward reference here
 

	
 
\subsubsection{User Products}
 

	
 
\label{user-product}
 

	
 
The scope of these requirements are narrow.  GPLv3~\S6 introduces the concept
 
of a ``User Product'', which includes devices that are sold for personal,
 
family, or household use.  Distributors are only required to provide
 
Installation Information when they convey object code in a User Product.
 

	
 
In brief, the right to convey object code in a defined class of ``User
 
Products,'' under certain circumstances, on providing whatever information is
 
required to enable a recipient to replace the object code with a functioning
 
modified version.
 

	
 
This was a compromise that was difficult for the FSF to agree to during the
 
GPLv3 drafting process.  However, companies and governments that use
 
specialized or enterprise-level computer facilities reported that they
 
actually \textit{want} their systems not to be under their own control.
 
Rather than agreeing to this as a concession, or bowing to pressure, they ask
 
for this as a \textit{preference}.  It is not clear that the GPL should interfere
 
here, since the main problem lies elsewhere.
 

	
 
While imposing technical barriers to modification is wrong regardless of
 
circumstances, the areas where restricted devices are of the greatest
 
practical concern today fall within the User Product definition.  Most, if
 
not all, technically-restricted devices running GPL-covered programs are
 
consumer electronics devices.  Moreover, the disparity in clout between the
 
manufacturers and these users makes it difficult for the users to reject
 
technical restrictions through their weak and unorganized market power.  Even
 
limited to User Products, this provision addresses the fundamental problem.
 

	
 
% FIXME-LATER: link \href to USC 2301
 

	
 
The core of the User Product definition is a subdefinition of ``consumer
 
product'' adapted from the Magnuson-Moss Warranty Act, a federal
 
consumer protection law in the USA found in 15~USC~\S2301: ``any tangible
 
personal property which is normally used for personal, family, or household
 
purposes.''  The USA has had three decades of experience of liberal
 
judicial and administrative interpretation of this definition in a manner
 
favorable to consumer rights.\footnote{The Magnuson-Moss consumer product
 
  definition itself has been influential in the USA and Canada, having been
 
  adopted in several state and provincial consumer protection laws.}
 
Ideally, this body of interpretation\footnote{The FSF, however, was very
 
  clear that incorporation of such legal interpretation was in no way
 
  intended work as a general choice of USA law for GPLv3.} will guide
 
interpretation of the consumer product subdefinition in GPLv3~\S6, and this
 
will hopefully provide a degree of legal certainty advantageous to device
 
manufacturers and downstream licensees alike.
 

	
 
One well-established interpretive principle under Magnuson-Moss is that
 
ambiguities are resolved in favor of coverage.  That is, in cases where
 
it is not clear whether a product falls under the definition of consumer
 
product, the product will be treated as a consumer product.\footnote{16
 
CFR~\S\ 700.1(a); \textit{McFadden v.~Dryvit Systems, Inc.}, 54
 
UCC~Rep.~Serv.2d 934 (D.~Ore.~2004).}  Moreover, for a given product,
 
``normally used'' is understood to refer to the typical use of that type
 
of product, rather than a particular use by a particular buyer.
 
Products that are commonly used for personal as well as commercial
 
purposes are consumer products, even if the person invoking rights is a
 
commercial entity intending to use the product for commercial
 
purposes.\footnote{16 CFR \S \ 700.1(a).  Numerous court decisions
 
interpreting Magnuson-Moss are in accord; see, e.g., \textit{Stroebner
 
Motors, Inc.~v.~Automobili Lamborghini S.p.A.}, 459 F.~Supp.2d 1028,
 
1033 (D.~Hawaii 2006).}  Even a small amount of ``normal'' personal use
 
is enough to cause an entire product line to be treated as a consumer
 
product under Magnuson-Moss\footnote{\textit{Tandy Corp.~v.~Marymac
 
Industries, Inc.}, 213 U.S.P.Q.~702 (S.D.~Tex.~1981). In this case, the
 
court concluded that TRS-80 microcomputers were consumer products, where
 
such computers were designed and advertised for a variety of users,
 
including small businesses and schools, and had only recently been
 
promoted for use in the home.}.
 

	
 
However, Magnuson-Moss is not a perfect fit because in the area of components
 
of dwellings, the settled interpretation under Magnuson-Moss under-inclusive.
 
Depending on how such components are manufactured or sold, they may or may
 
not be considered Magnuson-Moss consumer products.\footnote{Building
 
  materials that are purchased directly by a consumer from a retailer, for
 
  improving or modifying an existing dwelling, are consumer products under
 
  Magnuson-Moss, but building materials that are integral component parts of
 
  the structure of a dwelling at the time that the consumer buys the dwelling
 
  are not consumer products. 16 C.F.R.~\S\S~700.1(c)--(f); Federal Trade
 
  Commission, Final Action Concerning Review of Interpretations of
 
  Magnuson-Moss Warranty Act, 64 Fed.~Reg.~19,700 (April 22, 1999); see also,
 
  e.g., \textit{McFadden}, 54 U.C.C.~Rep.~Serv.2d at 934.}  Therefore, GPLv3
 
defines User Products as a superset of consumer products that also includes
 
``anything designed or sold for incorporation into a dwelling.''
 

	
 
Thus, the three sentences in the center of GPLv3's User Product definition
 
encapsulate the judicial and administrative principles established over the
 
past three decades in the USA concerning the Magnuson-Moss consumer product
 
definition.  First, it states that doubtful cases are resolved in favor of
 
coverage under the definition.  Second, it indicate that the words ``normally
 
used'' in the consumer product definition refer to a typical or common use of
 
a class of product, and not the status of a particular user or expected or
 
actual uses by a particular user.  Third, it clearly states that the
 
existence of substantial non-consumer uses of a product does not negate a
 
determination that it is a consumer product, unless such non-consumer uses
 
represent the only significant mode of use of that product.
 

	
 
It should be clear from these added sentences that it is the general mode of
 
use of a product that determines objectively whether or not it is a consumer
 
product.  One could not escape the effects of the User Products provisions by
 
labeling what is demonstrably a consumer product in ways that suggest it is
 
``for professionals'', for example.
 

	
 

	
 
\subsubsection{Installation Information}
 

	
 
With the User Products definition complete,  The ``Installation Information''
 
definition uses that to define what those receiving object code inside a User
 
Product must receive.
 

	
 
Installation Information is information that is ``required to install and
 
execute modified versions of a covered work \dots from a modified version of
 
its'' CCS, in the same User Product for which the covered work is conveyed.
 
GPLv3 provides guidance concerning how much information must be provided: it
 
``must suffice to ensure that the continued functioning of the modified
 
object code is in no case prevented or interfered with solely because
 
modification has been made.''  For example, the information provided would be
 
insufficient if it enabled a modified version to run only in a disabled
 
fashion, solely because of the fact of modification (regardless of the actual
 
nature of the modification).  The information need not consist of
 
cryptographic keys; Installation Information may be ``any methods,
 
procedures, authorization keys, or other information''.
 

	
 
Note that GPLv3 does not define ``continued functioning'' further.  However,
 
GPLv3 does provide some additional guidance concerning the scope of
 
GPLv3-compliant action or inaction that distributors of
 
technically-restricted User Products can take with respect to a downstream
 
recipient who replaces the conveyed object code with a modified version.
 
First of all, GPLv3 makes clear that GPLv3 implies no obligation ``to
 
continue to provide support service, warranty, or updates'' for such a work.
 

	
 
Second, most technically-restricted User Products are designed to communicate
 
across networks.  It is important for both users and network providers to
 
know when denial of network access to devices running modified versions
 
becomes a GPL violation.  GPLv3 permits denial of access in two cases: ``when
 
the modification itself materially and adversely affects the operation of the
 
network,'' and when the modification itself ``violates the rules and
 
protocols for communication across the network''.  The second case is
 
deliberately drawn in general terms, and it serves as a foundation for
 
reasonable enforcement policies that respect recipients' right to modify
 
while recognizing the legitimate interests of network providers.
 

	
 
Note that GPLv3 permits the practice of conveying object code in a mode not
 
practically susceptible to modification by any party, such as code burned in
 
ROM or embedded in silicon.  The goal of the Installation Information
 
requirement is to ensure the downstream licensee receives the real right to
 
modify when the device manufacturer or some other party retains that right.
 
Accordingly, GPLv3\S6's ante-penultimate paragraph states that the
 
requirement to provide Installation Information ``does not apply if neither
 
you nor any third party retains the ability to install modified object code
 
on the User Product''.
 

	
 
Finally, GPLv3\S6 makes it clear that there is also no requirement to
 
provide warranty or support for the User Product itself.
 

	
 
\subsection{GPLv3~\S7: Additional Permissions}
 
\label{GPLv3s7}
 

	
 
The GPL is a statement of permissions, some of which have conditions.
 
Additional terms --- terms that supplement those of the GPL --- may come to be
 
placed on, or removed from, GPL-covered code in certain common ways.
 
Copyleft licensing theorists have generally called
 
 those added terms ``additional permissions'' if they grant
 
exceptions from the conditions of the GPL, and ``additional requirements'' if
 
they add conditions to the basic permissions of the GPL\@. The treatment of
 
additional permissions and additional requirements under GPLv3 is necessarily
 
asymmetrical, because they do not raise the same interpretive
 
issues; in particular, additional requirements, if allowed without careful
 
limitation, could transform a GPL'd program into a non-free one.
 

	
 
With these principles in the background, GPLv3~\S7  answers the following
 
questions: 
 
\begin{enumerate}
 
\item How does the presence of additional terms on all or part of a GPL'd program
 
affect users' rights?
 

	
 
\item When and how may a licensee add terms to code being
 
distributed under the GPL? 
 

	
 
\item When may a licensee remove additional terms?
 
\end{enumerate}
 

	
 
Additional permissions present the easier case.  Since the mid-1990s,
 
permissive exceptions often appeared alongside GPLv2 to allow combination
 
with certain non-free code.  Typically, downstream
 
stream recipients could remove those exceptions and operate under pure GPLv2.
 
Similarly, LGPLv2.1 is in essence a permissive variant of GPLv2,
 
and it permits relicensing under the GPL\@.  
 

	
 
These practices are now generalized via GPLv3~\S7.
 
A licensee may remove any additional permission from
 
a covered work, whether it was placed by the original author or by an
 
upstream distributor.  A licensee may also add any kind of additional
 
permission to any part of a work for which the licensee has, or can give,
 
appropriate copyright permission. For example, if the licensee has written
 
that part, the licensee is the copyright holder for that part and can
 
therefore give additional permissions that are applicable to it.
 
Alternatively, the part may have been written by someone else and licensed,
 
with the additional permissions, to that licensee.  Any additional
 
permissions on that part are, in turn, removable by downstream recipients.
 
As GPLv3~\S7\P1 explains, the effect of an additional permission depends on
 
whether the permission applies to the whole work or a part.
 

	
 
% FIXME-LATER: LGPLv3 will have its own section
 

	
 
Indeed, LGPLv3 is itself simply  a list of additional permissions supplementing the
 
terms of GPLv3.  GPLv3\S7 has thus provided the basis for recasting a
 
formally complex license as an elegant set of added terms, without changing
 
any of the fundamental features of the existing LGPL\@.  LGPLv3 is thus  a model for developers wishing to license their works under the
 
GPL with permissive exceptions.  The removability of additional permissions
 
under GPLv3\S7 does not alter any existing behavior of the LGPL since the LGPL
 
has always allowed relicensing under the ordinary GPL\@.
 

	
 
\section{GPLv3~\S7: Understanding License Compatibility}
 
\label{license-compatibility}
 

	
 
A challenge that faced the Free Software community heavily through out the
 
early 2000s was the proliferation of incompatible Free Software licenses.  Of
 
course, the GPL cannot possibly be compatible with all such licenses.
 
However, GPLv3
 
contains provisions that are designed to reduce license incompatibility by
 
making it easier for developers to combine code carrying non-GPL terms with
 
GPL'd code.
 

	
 
This license compatibility issue arises for
 
three reasons.  First, the GPL is a strong copyleft license, requiring
 
modified versions to be distributed under the GPL\@.  Second, the GPL states
 
that no further restrictions may be placed on the rights of recipients.
 
Third, all other software freedom respecting licenses in common use contain certain
 
requirements, many of which are not conditions made by the GPL\@.  Thus, when
 
GPL'd code is modified by combination with code covered by another formal
 
license that specifies other requirements, and that modified code is then
 
distributed to others, the freedom of recipients may be burdened by
 
additional requirements in violation of the GPL.  It can be seen that
 
additional permissions in other licenses do not raise any problems of license
 
compatibility.
 

	
 
GPLv3  took a new approach to the issue of combining GPL'd code with
 
code governed by the terms of other software freedom licenses.  Traditional
 
GPLv2 license compatibility theory (which was not explicitly stated in GPLv2
 
itself, but treated as a license interpretation matter by the FSF) held that GPLv2 allowed such
 
combinations only if the non-GPL licensing terms permitted distribution under
 
the GPL and imposed no restrictions on the code that were not also imposed by
 
the GPL\@.  In practice, the FSF historically supplemented that policy with a structure of
 
exceptions for certain kinds of combinations.
 

	
 
GPLv3~\S7  implements a more explicit policy on license
 
compatibility.  It formalizes the circumstances under which a licensee may
 
release a covered work that includes an added part carrying non-GPL terms. 
 
GPLv3~\S7 distinguish between terms that provide additional permissions, and terms that
 
place additional requirements on the code, relative to the permissions and
 
requirements established by applying the GPL to the code.
 

	
 
As discussed in the previous section of this tutorial, GPLv3~\S7 first and foremost explicitly allows added parts covered by terms with
 
additional permissions to be combined with GPL'd code. This codifies the
 
existing practice of regarding such licensing terms as compatible with the
 
GPL\@. A downstream user of a combined GPL'd work who modifies such an added
 
part may remove the additional permissions, in which case the broader
 
permissions no longer apply to the modified version, and only the terms of
 
the GPL apply to it.
 

	
 
In its treatment of terms that impose additional requirements, GPLv3\S7
 
extends the range of licensing terms with which the GPL is compatible.  An
 
added part carrying additional requirements may be combined with GPL'd code,
 
but only if those requirements belong to an set enumerated in GPLv3\S7. There
 
are, of course,  limits on the acceptable additional requirements, which 
 
ensures that enhanced license compatibility does not
 
defeat the broader software-freedom-defending terms of the GPL\@. Unlike terms that grant
 
additional permissions, terms that impose additional requirements cannot be
 
removed by a downstream user of the combined GPL'd work, because only in the
 
pathological case\footnote{Theoretically, a user could collect copyright
 
  assignment from all known contributors and then do this, but this would
 
  indeed be the pathological case.}  would a user have the right to do so.
 

	
 
% FIXME-LATER: It would be good to have detailed info on each of 7a-f.
 
%              Here's some commented-out text that might be useful for 7a-b
 

	
 
%% Under subsections 7a and 7b, the requirements may include preservation of
 
%% copyright notices, information about the origins of the code or alterations
 
%% of the code, and different warranty disclaimers. Under subsection 7c, the
 
%% requirements may include limitations on the use of names of contributors and
 
%% on the use of trademarks for publicity purposes. In general, we permit these
 
%% requirements in added terms because many free software licenses include them
 
%% and we consider them to be unobjectionable. Because we support trademark fair
 
%% use, the limitations on the use of trademarks may seek to enforce only what
 
%% is required by trademark law, and may not prohibit what would constitute fair
 
%% use.
 

	
 
% FIXME-LATER:  Say removing additional restrictions
 

	
 
% FIXME-LATER: This text may be useful later:
 

	
 
%% Some have questioned whether section 7 is needed, and some have suggested
 
%% that it creates complexity that did not previously exist.  We point out to
 
%% those readers that there is already GPLv2-licensed code that carries
 
%% additional terms.  One of the objectives of section 7 is to rationalize
 
%% existing practices of program authors and modifiers by setting clear
 
%% guidelines regarding the removal and addition of such terms.  With its
 
%% carefully limited list of allowed additional requirements, section 7
 
%% accomplishes additional objectives, permitting the expansion of the base of
 
%% code available for GPL developers, while also encouraging useful
 
%% experimentation with requirements we do not include in the GPL itself.
 

	
 
\section{GPLv3~\S8: A Lighter Termination}
 

	
 
GPLv2 provided for automatic termination of the rights of a person who
 
copied, modified, sublicensed, or distributed a work in violation of the
 
license.  Automatic termination can be too harsh for those who have committed
 
an inadvertent violation, particularly in cases involving distribution of
 
large collections of software having numerous copyright holders.  A violator
 
who resumes compliance with GPLv2 technically needs to obtain forgiveness
 
from all copyright holders, and even contacting them all might be impossible.
 

	
 
GPLv3~\S8 now grants opportunities for provisional and permanent
 
reinstatement of rights. The termination procedure provides a limited
 
opportunity to cure license violations.  If a licensee has committed a
 
first-time violation of the GPL with respect to a given copyright holder, but
 
the licensee cures the violation within 30 days following receipt of notice
 
of the violation, then any of the licensee's GPL rights that have been
 
terminated by the copyright holder are ``automatically reinstated''.
 

	
 

	
 
Finally, if a licensee violates the GPL, a contributor may terminate any
 
patent licenses that it granted under GPLv3~\S11, in addition to any
 
copyright permissions the contributor granted to the licensee.
 

	
 
% FIXME-LATER: write more here, perhaps linking up to enforcement
 

	
 
\section{GPLv3~\S9: Acceptance}
 

	
 
GPLv3~\S9 means what it says: mere receipt or execution of code neither
 
requires nor signifies contractual acceptance under the GPL.  Speaking more
 
broadly, GPLv3 is intentionally structured as a unilateral grant
 
of copyright permissions, the basic operation of which exists outside of any
 
law of contract.  Whether and when a contractual relationship is formed
 
between licensor and licensee under local law do not necessarily matter to
 
the working of the license.
 

	
 
\section{GPLv3~\S10: Explicit Downstream License}
 
\label{GPLv3s10}
 

	
 
% FIXME-LATER: this is a punt: need more time to write!
 

	
 
GPLv3~\S10 ensures that everyone downstream receives licenses from all
 
copyright holders.  It really is a generally straightforward section.
 

	
 
% FIXME-LATER: link up this paragraph to above sections.
 

	
 
Note, however, GPLv3 removed the words ``at no charge'' from GPLv2~\S2(b) (in
 
GPLv3,~\S5(b)) because it contributed to a misconception that the GPL did not
 
permit charging for distribution of copies.  The purpose of the ``at no
 
charge'' wording was to prevent attempts to collect royalties from third
 
parties.  The removal of these words created the danger that the imposition
 
of licensing fees would no longer be seen as a license violation.  Therefore,
 
GPLv3~\S10 adds a new explicit prohibition on imposition of licensing fees or
 
royalties.  This section is an appropriate place for such a clause, since it
 
is a specific consequence of the general requirement that no further
 
restrictions be imposed on downstream recipients of GPL-covered code.
 

	
 
% FIXME-LATER: This text needs further study before I can conclude it belongs
 
% in this tutorial:
 

	
 
%% Careful readers of the GPL have suggested that its explicit prohibition
 
%% against imposition of further restrictions\footnote{GPLv2, section 6; Draft
 
%%   3, section 10, third paragraph.} has, or ought to have, implications for
 
%% those who assert patents against other licensees.  Draft 2 took some steps to
 
%% clarify this point in a manner not specific to patents, by describing the
 
%% imposition of ``a license fee, royalty, or other charge'' for exercising GPL
 
%% rights as one example of an impermissible further restriction.  In Draft 3 we
 
%% have clarified further that the requirement of non-imposition of further
 
%% restrictions has specific consequences for litigation accusing GPL-covered
 
%% programs of infringement.  Section 10 now states that ``you may not initiate
 
%% litigation (including a cross-claim or counterclaim in a lawsuit) alleging
 
%% that any patent claim is infringed by making, using, selling, offering for
 
%% sale, or importing the Program (or the contribution of any contributor).''
 
%% That is to say, a patent holder's licensed permissions to use a work under
 
%% GPLv3 may be terminated under section 8 if the patent holder files a lawsuit
 
%% alleging that use of the work, or of any upstream GPLv3-licensed work on
 
%% which the work is based, infringes a patent.
 

	
 
\section{GPLv3~\S11: Explicit Patent Licensing}
 
\label{GPLv3s11}
 

	
 
Software patenting is a harmful and unjust policy, and should be abolished;
 
recent experience makes this all the more evident. Since many countries grant
 
patents that can apply to and prohibit software packages, in various guises
 
and to varying degrees, GPLv3 seeks to protect the users of GPL-covered programs
 
from those patents, while at the same time making it feasible for patent
 
holders to contribute to and distribute GPL-covered programs as long as they
 
do not attack the users of those programs.
 

	
 
It is generally understood that GPLv2 implies some limits on a licensee's
 
power to assert patent claims against the use of GPL-covered works.
 
However, the patent licensing practices that GPLv2~\S7 (corresponding to
 
GPLv3~\S12) is designed to prevent is only one of several ways in which
 
software patents threaten to make free programs non-free and to prevent users
 
from exercising their rights under the GPL. GPLv3 takes a more comprehensive
 
approach to combating the danger of patents.
 

	
 
GPLv2~\S7 has seen some success in deterring conduct that would otherwise
 
result in denial of full downstream enjoyment of GPL rights, and thus it is
 
preserved in GPLv3~\S12.  Experience has shown that more is necessary,
 
however, to ensure adequate community safety where companies act in concert
 
to heighten the anticompetitive use of patents that they hold or license.
 

	
 
Therefore,  GPLv3 is designed to reduce the patent risks that distort and
 
threaten the activities of users who make, run, modify and share Free
 
Software.  At the same time, GPLv3 gives favorable consideration to practical
 
goals such as certainty and administrability for patent holders that
 
participate in distribution and development of GPL-covered software.  GPLv3's
 
policy requires each such patent holder to provide appropriate levels of
 
patent assurance to users, according to the nature of the patent holder's
 
relationship to the program.
 

	
 
\subsection{The Contributor's Explicit Patent License}
 

	
 
Specifically, the ideal might have been for GPLv3 to feature a patent license
 
grant triggered by all acts of distribution of GPLv3-covered works.  The FSF
 
considered it during the GPLv3 drafting process, but many patent-holding
 
companies objected to this policy.  They have made two objections: (1) the
 
far-reaching impact of the patent license grant on the patent holder is
 
disproportionate to the act of merely distributing code without modification
 
or transformation, and (2) it is unreasonable to expect an owner of vast
 
patent assets to exercise requisite diligence in reviewing all the
 
GPL-covered software that it provides to others.  Some expressed particular
 
concern about the consequences of ``inadvertent'' distribution.
 

	
 
The argument that the impact of the patent license grant would be
 
``disproportionate'',  that is to say unfair, is not valid. Since
 
software patents are weapons that no one should have, and using them for
 
aggression against free software developers is an egregious act (thus
 
preventing that act cannot be unfair). 
 

	
 
However, the second argument seems valid in a practical sense.  A
 
typical GNU/Linux distribution includes thousands of programs.  It would
 
be quite difficult for a re-distributor with a large patent portfolio to
 
review all those programs against that portfolio every time it receives
 
and passes on a new version of the distribution.  Moreover, this question
 
raises a strategic issue. If the GPLv3 patent license requirements
 
convince patent-holding companies to remain outside the distribution
 
path of all GPL-covered software, then these requirements, no matter how
 
strong, will cover few patents. 
 

	
 
GPLv3 therefore makes a partial concession
 
which would lead these companies to feel secure in doing the
 
distribution themselves. GPLv3~\S11
 
applies only to those distributors that have
 
modified the program.  The other changes we have made in sections 10 and
 
11 provide strengthened defenses against patent assertion and compensate
 
partly for this concession. 
 

	
 
Therefore, GPLv3~\S11 introduces the terms ``contributor'', ``contributor version'', and
 
``essential patent claims'', which are
 
used in the GPLv3~\S11\P3.   Viewed from the perspective of a recipient of the
 
Program, contributors include all the copyright holders for the Program,
 
other than copyright holders of material originally licensed under non-GPL
 
terms and later incorporated into a GPL-covered work.  The contributors are
 
therefore the initial GPLv3 licensors of the Program and all subsequent
 
upstream licensors who convey, under the terms of GPLv3~\S5, modified covered
 
works.
 
Thus, the ``contributor version'' includes the material the contributor has copied from the
 
upstream version that the contributor has modified.  GPLv3~\S11\P3
 
 does not apply to those that redistribute the program
 
without change.\footnote{An implied patent license from the distributor,
 
however, often arises.  See \S~\ref{gpl-implied-patent-grant} in this tutorial}
 
In other words, the ``contributor version'' includes not just
 
the material added or altered by the contributor, but also the pre-existing
 
material the contributor copied from the upstream version and retained in the
 
modified version.  (GPLv3's usage of ``contributor'' and ``contribution'' should
 
not be confused with the various other ways in which those terms are used in
 
certain other free software licenses\footnote{Cf., e.g., Apache License,
 
  version 2.0, section 1; Eclipse Public License, version 1.0, section 1;
 
  Mozilla Public License, version 1.1, section 1.1.}.)
 

	
 
Some details of the ``essential patent claims'' definition deserve special
 
mention.  ``Essential patent claims'', for a given party, are a subset of the
 
claims ``owned or controlled'' by the party.  They do include sublicensable
 
claims that have been licensed to the contributor by a third
 
party.\footnote{This issue is typically handled in other software freedom
 
  licenses having patent licensing provisions by use of the unhelpful term
 
  ``licensable,'' which is either left undefined or is given an ambiguous
 
  definition.}  Most commercial patent license agreements that permit
 
sublicensing do so under restrictive terms that are inconsistent with the
 
requirements of the GPL\@.  For example, some patent licenses allow the
 
patent licensee to sublicense but require collection of royalties from any
 
sublicensees.  The patent licensee could not distribute a GPL-covered program
 
and grant the recipient a patent sublicense for the program without violating
 
section 12 of GPLv3.\footnote{GPLv3 also provides an example in section 12
 
  that makes this point clear.}  In rare cases, however, a conveying party
 
can freely grant patent sublicenses to downstream recipients without
 
violating the GPL\@.
 

	
 
Additionally, ``essential patent claims'' are those patents ``that would be
 
infringed by some manner, permitted by this License, of making, using, or
 
selling the work''.  This intends to make clear that a patent claim is
 
``essential'' if some mode of usage would infringe that claim, even if there
 
are other modes of usage that would not infringe.
 

	
 
Finally, ``essential patent claims \ldots do not include
 
claims that would be infringed only as a consequence of further
 
modification of the work.''  The set of essential patent
 
claims licensed  is fixed by the
 
particular version of the work that was contributed.  The claim set
 
cannot expand as a work is further modified downstream.  (If it could,
 
then any software patent claim would be included, since any software
 
patent claim can be infringed by some further modification of the
 
work.)\footnote{However, ``the work'' should not be understood to be
 
restricted to a particular mechanical affixation of, or medium for
 
distributing, a program, where the same program might be provided in
 
other forms or in other ways that may be captured by other patent claims
 
held by the contributor.}
 

	
 
\medskip
 

	
 
Ideally, this contributor patent policy will result in fairly frequent licensing of patent
 
claims by contributors.  A contributor is charged with awareness of the fact
 
that it has modified a work and provided it to others; no act of contribution
 
should be treated as inadvertent.  GPLv3's rule also requires no more work, for a
 
contributor, than the weaker rule proposed by the patent holders.  Under
 
their rule, the contributor must always compare the entire work against its
 
patent portfolio to determine whether the combination of the modifications
 
with the remainder of the work cause it to read on any of the contributor's
 
patent claims.
 

	
 
\subsection{Conveyors' Patent Licensing}
 

	
 
The remaining patent licensing in GPLv3 deals with patent licenses that are
 
granted by conveyance.  The licensing is not as complete or far reaching as
 
the contributor patent licenses discussed in the preceding section.
 

	
 
The term ``patent license,'' as used in GPLv3~\S11\P4--6, is not meant to be
 
confined to agreements formally identified or classified as patent licenses.
 
GPLv3~\S11\P3  makes this clear by defining ``patent
 
license,'' for purposes of the subsequent three paragraphs, as ``any express
 
agreement or commitment, however denominated, not to enforce a patent
 
(such as an express permission to practice a patent or covenant not to
 
sue for patent infringement)''
 

	
 
% FIME-LATER: I want to ask Fontana about this before adding it.
 
% FIXME-LATER: I want to ask Fontana about this before adding it.
 

	
 
% The definition does not include patent licenses that arise by
 
% implication or operation of law, because the third through fifth paragraphs
 
% of section 11 are specifically concerned with explicit promises that purport
 
% to be legally enforceable.
 

	
 
GPLv3~\S11\P5 is commonly called GPLv3's downstream shielding provision.  It
 
responds particularly to the problem of exclusive deals between patent
 
holders and distributors, which threaten to distort the free software
 
distribution system in a manner adverse to developers and users.  The
 
fundamental idea is to make a trade-off between assuring a patent license for
 
downstream and making  (possibly patent-encumbered) CCS publicly available.
 

	
 
Simply put, in nearly all cases in which the ``knowingly relying'' test is
 
met, the patent license will indeed not be sublicensable or generally
 
available to all on free terms.  If, on the other hand, the patent license is
 
generally available under terms consistent with the requirements of the GPL,
 
the distributor is automatically in compliance, because the patent license
 
has already been extended to all downstream recipients.  Finally, if the
 
patent license is sublicensable on GPL-consistent terms, the distributor may
 
choose to grant sublicenses to downstream recipients instead of causing the
 
CCS to be publicly available.  (In such a case, if the distributor is also a
 
contributor, it will already have granted a patent sublicense anyway, and so
 
it need not do anything further to comply with the third paragraph.)
 

	
 
Admittedly, public disclosure of CCS is not necessarily required by other
 
sections of the GPL, and the FSF in drafting GPLv3 did not necessarily wish
 
to impose a general requirement to make source code available to all, which
 
has never been a GPL condition.  However, many vendors who produce products
 
that include copylefted software, and who are most likely to be affected by the
 
downstream shielding provision, lobbied for the addition of the source code
 
availability option, so it remains.
 

	
 
Meanwhile, two specific alternatives to the source code availability option
 
are also available. The distributor may comply by disclaiming the patent
 
license it has been granted for the conveyed work, or by arranging to extend
 
the patent license to downstream recipients\footnote{The latter option, if
 
  chosen, must be done ``in a manner consistent with the requirements of this
 
  License''; for example, it is unavailable if extension of the patent
 
  license would result in a violation of GPLv3~\S 12.}.  The GPL is intended
 
to permit private distribution as well as public distribution, and the
 
addition of these options ensures that this remains the case, even though it
 
remains likely that distributors in this situation will usually choose the
 
source code availability option.
 

	
 
Note that GPLv3~\S11\P5 is activated only if the CCS is not already otherwise
 
publicly available.  (Most often it will, in fact, already be available on
 
some network server operated by a third party.)  Even if it is not already
 
available, the option to ``cause the Corresponding Source to be so
 
available'' can then be satisfied by verifying that a third party has acted
 
to make it available.  That is to say, the affected distributor need not
 
itself host the CCS to take advantage of the source code availability option.
 
This subtlety may help the distributor avoid certain peculiar assumptions of
 
liability.
 

	
 
Note that GPLv3~\S11\P6--7 are designed to stop distributors from colluding with
 
third parties to offer selective patent protection.  GPLv3 is designed to
 
ensure that all users receive the same rights; arrangements that circumvent
 
this make a mockery of free software, and we must do everything in our power
 
to stop them.
 

	
 
First, GPLv3~\S11\P6 states that any license that protects some recipients of
 
GPL'd software must be extended to all recipients of the software.  
 
If conveyors arrange to provide patent
 
protection to some of the people who get the software from you, that
 
protection is automatically extended to everyone who receives the software,
 
no matter how they get it. 
 

	
 
Second, GPLv3~\S11\P7
 
prohibit anyone who made such an agreement from distributing software
 
released under GPLv3.    Conveyors are prohibited from
 
distributing software under GPLv3 if the conveyor makes an agreement of that
 
nature in the future.
 

	
 
The date in GPLv3~\S11\P7 likely seems arbitrary to those who did not follow
 
the GPLv3 drafting process.  This issue was hotly debated during the drafting of
 
GPLv3, but ultimately one specific deal of this type --- a deal between Microsoft
 
and Novell for Microsoft to provide so-called ``coupons'' to Microsoft customers to redeem
 
for copies of Novell's GNU/Linux distribution with a Microsoft patent license -- was
 
designed to be excluded.
 

	
 
The main reason for this was a tactical decision by the FSF.  FSF believed they can do more to
 
protect the community by allowing Novell to use software under GPLv3
 
than by forbidding it to do so.  This is because of
 
paragraph 6 of section 11 (corresponding to paragraph 4 in Draft 3).
 
It will apply, under the Microsoft/Novell deal, because of the coupons
 
that Microsoft has acquired that essentially commit it to participate
 
in the distribution of the Novell SLES GNU/Linux system.
 

	
 
The FSF also gave a secondary reason:  to avoid affecting other kinds of agreements for
 
other kinds of activities.  While GPLv3 sought to 
 
distinguish pernicious deals of the Microsoft/Novell type from
 
business conduct that is not particularly harmful, the FSF also did not
 
assume success in that drafting, and thus there remained some risk that other
 
unchangeable past agreements could fall within the  scope of GPLv3~\S11\P7.
 
In future deals, distributors engaging in ordinary business practices
 
can structure the agreements so that they do not fall under GPLv3~\S11\P7.
 

	
 
\section{GPLv3~\S12: Familiar as GPLv2~\S7}
 

	
 
GPLv2~\S12 remains almost completely unchanged from the text that appears in
 
GPLv2~\S7.  This is an important provision that ensures a catch-all to ensure
 
that nothing ``surprising'' interferes with the continued conveyance safely
 
under copyleft.
 

	
 
The wording in the first sentence of GPLv3~\S12 has been revised slightly to
 
clarify that an agreement -- such as a litigation settlement agreement or a
 
patent license agreement -- is one of the ways in which conditions may be
 
``imposed'' on a GPL licensee that may contradict the conditions of the GPL,
 
but which do not excuse the licensee from compliance with those conditions.
 
This change codifies the historical interpretation of GPLv2.
 

	
 
GPLv3 removed the limited severability clause of GPLv2~\S7 as a
 
matter of tactical judgment, believing that this is the best way to ensure
 
that all provisions of the GPL will be upheld in court. GPLv3 also removed
 
the final sentence of GPLv2 section 7, which the FSF consider to be unnecessary.
 

	
 
\section{GPLv3~\S13: The Great Affero Compromise}
 

	
 
The Affero GPL was written with the expectation that its
 
additional requirement would be incorporated into the terms of GPLv3
 
itself.  Many software freedom advocates, including some authors of this
 
tutorial, advocated heavily for that, and fully expected it to happen.
 

	
 
The FSF, however, chose not to include the Affero clause in GPLv3, due to
 
what it called  ``irreconcilable views from
 
different parts of the community''.  Many
 
commercial users of Free Software were opposed to the inclusion of a
 
mandatory Affero-like requirement in the body of GPLv3 itself.  In fact, some
 
wealthier companies even threatened to permanently fund forks of many FSF
 
copyrighted-programs under GPLv2 if the Affero clause appeared in GPLv3.
 

	
 
Meanwhile, there was disagreement even among copyleft enthusiasts about the
 
importance of the provision.  A coalition never formed, and ultimately the
 
more powerful interest implicitly allied with the companies who deeply opposed
 
the Affero clause such that the FSF felt the Affero clause would need its own
 
license, but one compatible with GPLv3. 
 

	
 
GPLv3~\S13 makes GPLv3 compatible with the AGPLv3, so that at least code can
 
be shared between AGPLv3'd and GPLv3'd projects, even if the Affero clause
 
does not automatically apply to all GPLv3'd works.
 

	
 
%FIXME-LATER:  no time to do this justice, will come back later, instead the
 
%above.
 

	
 
%% Some of this hostility seemed to be based on a misapprehension that
 
%% Affero-like terms placed on part of a covered work would somehow extend
 
%% to the whole of the work.\footnote{It is possible that the presence of
 
%% the GPLv2-derived copyleft clause in the existing Affero GPL contributed
 
%% to this misunderstanding.}  Our explanations to the contrary did little
 
%% to satisfy these critics; their objections to 7b4 instead evolved into a
 
%% broader indictment of the additional requirements scheme of section 7.
 
%% It was clear, however, that much of the concern about 7b4 stemmed from
 
%% its general formulation.  Many were alarmed at the prospect of GPLv3
 
%% compatibility for numerous Affero-like licensing conditions,
 
%% unpredictable in their details but potentially having significant
 
%% commercial consequences.
 

	
 
%% On the other hand, many developers, otherwise sympathetic to the policy
 
%% goals of the Affero GPL, have objected to the form of the additional
 
%% requirement in that license.  These developers were generally
 
%% disappointed with our decision to allow Affero-like terms through
 
%% section 7, rather than adopt a condition for GPLv3.  Echoing their
 
%% concerns about the Affero GPL itself, they found fault with the wording
 
%% of the section 7 clause in both of the earlier drafts.  We drafted 7b4
 
%% at a higher level than its Draft 1 counterpart based in part on comments
 
%% from these developers. They considered the Draft 1 clause too closely
 
%% tied to the Affero mechanism of preserving functioning facilities for
 
%% downloading source, which they found too restrictive of the right of
 
%% modification.  The 7b4 rewording did not satisfy them, however. They
 
%% objected to its limitation to terms requiring compliance by network
 
%% transmission of source, and to the technically imprecise or inaccurate
 
%% use of the phrase ``same network session.''
 

	
 
%% We have concluded that any redrafting of the 7b4 clause would fail to
 
%% satisfy the concerns of both sets of its critics.  The first group
 
%% maintains that GPLv3 should do nothing about the problem of public
 
%% use. The second group would prefer for GPLv3 itself to have an
 
%% Affero-like condition, but that seems to us too drastic. By permitting
 
%% GPLv3-covered code to be linked with code covered by version 2 of the
 
%% Affero GPL, the new section 13 honors our original commitment to
 
%% achieving GPL compatibility for the Affero license.
 

	
 
%% Version 2 of the Affero GPL is not yet published.  We will work with
 
%% Affero, Inc., and with all other interested members of our community, to
 
%% complete the drafting of this license following the release of Draft 3,
 
%% with a goal of having a final version available by the time of our
 
%% adoption of the final version of GPLv3.  We hope the new Affero license
 
%% will satisfy those developers who are concerned about the issue of
 
%% public use of unconveyed versions but who have concerns about the
 
%% narrowness of the condition in the existing Affero license.
 

	
 
%% As the second sentence in section 13 indicates, when a combined work is
 
%% made by linking GPLv3-covered code with Affero-covered code, the
 
%% copyleft on one part will not extend to the other part.\footnote{The
 
%% plan is that the additional requirement of the new Affero license will
 
%% state a reciprocal limitation.} That is to say, in such combinations,
 
%% the Affero requirement will apply only to the part that was brought into
 
%% the combination under the Affero license.  Those who receive such a
 
%% combination and do not wish to use code under the Affero requirement may
 
%% remove the Affero-covered portion of the combination.
 

	
 
Meanwhile, those who criticize the permission to link with code under the Affero
 
GPL should recognize that most other free software licenses also permit
 
such linking. 
 

	
 
\section{GPLv3~\S14: So, When's GPLv4?}
 
\label{GPLv3s14}
 

	
 
No substantive change has been made in section 14. The wording of the section
 
has been revised slightly to make it clearer.
 

	
 
It's unclear when the FSF might consider publishing GPLv4.  However, this
 
section makes it clear that the FSF is the sole authority who can decide
 
such.
 

	
 
The main addition to this section allows a third-party proxy to be appointed
 
by contributors who wish someone else to make relicensing to new versions of
 
GPL when they are released.  This is a ``halfway'' point between using ``-only''
 
or ``-or-later'' by consolidating the decision-making on that issue to a
 
single authority.
 

	
 
% FIXME-LATER: better proxy description
 

	
 
\section{GPLv3~\S15--17: Warranty Disclaimers and Liability Limitation}
 

	
 
No substantive changes have been made in sections 15 and 16.
 

	
 
% FIXME-LATER: more, plus 17
 

	
 
% FIXME-LATER: Section header needed here about choice of law.
 

	
 
% FIXME-LATER: reword into tutorial
 

	
 
%% Some have asked us to address the difficulties of internationalization
 
%% by including, or permitting the inclusion of, a choice of law
 
%% provision.  We maintain that this is the wrong approach.  Free
 
%% software licenses should not contain choice of law clauses, for both
 
%% legal and pragmatic reasons.  Choice of law clauses are creatures of
 
%% contract, but the substantive rights granted by the GPL are defined
 
%% under applicable local copyright law. Contractual free software
 
%% licenses can operate only to diminish these rights.  Choice of law
 
%% clauses also raise complex questions of interpretation when works of
 
%% software are created by combination and extension.  There is also the
 
%% real danger that a choice of law clause will specify a jurisdiction
 
%% that is hostile to free software principles.
 

	
 
%% % FIXME-LATER: reword into tutorial, \ref to section 7.
 

	
 
%% Our revised version of section 7 makes explicit our view that the
 
%% inclusion of a choice of law clause by a licensee is the imposition of
 
%% an additional requirement in violation of the GPL.  Moreover, if a
 
%% program author or copyright holder purports to supplement the GPL with
 
%% a choice of law clause, section 7 now permits any licensee to remove
 
%% that clause.
 

	
 

	
 
% FIXME-LATER: does this need to be a section, describing how it was out then in
 
% then out then in? :)
 

	
 
Finally, the FSF shortened the section on ``How to Apply These
 
Terms to Your New Programs'' to just the bare essentials.
 

	
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 
\chapter{The Lesser GPL}
 

	
 
As we have seen in our consideration of the GPL, its text is specifically
 
designed to cover all possible derivative works under copyright law. Our
 
goal in designing the GPL was to make sure that any derivative work of GPL'd
 
software was itself released under the GPL when distributed. Reaching as far
 
as copyright law will allow is the most direct way to reach that goal.
 

	
 
However, while the strategic goal is to bring as much Free Software
 
into the world as possible, particular tactical considerations
 
regarding software freedom dictate different means. Extending the
 
copyleft effect as far as copyright law allows is not always the most
 
prudent course in reaching the goal. In particular situations, even
 
those of us with the goal of building a world where all published
 
software is Free Software realize that full copyleft does not best
 
serve us. The GNU Lesser General Public License (``GNU LGPL'') was
 
designed as a solution for such situations.
 

	
 
\section{The First LGPL'd Program}
 

	
 
The first example that FSF encountered where such altered tactics were
 
needed was when work began on the GNU C Library. The GNU C Library would
 
become (and today, now is) a drop-in replacement for existing C libraries.
 
On a Unix-like operating system, C is the lingua franca and the C library
 
is an essential component for all programs. It is extremely difficult to
 
construct a program that will run with ease on a Unix-like operating
 
system without making use of services provided by the C library --- even
 
if the program is written in a language other than C\@. Effectively, all
 
user application programs that run on any modern Unix-like system must
 
make use of the C library.
 

	
 
By the time work began on the GNU implementation of the C libraries, there
 
were already many C libraries in existence from a variety of vendors.
 
Every proprietary Unix vendor had one, and many third parties produced
 
smaller versions for special purpose use. However, our goal was to create
 
a C library that would provide equivalent functionality to these other C
 
libraries on a Free Software operating system (which in fact happens today
 
on modern GNU/Linux systems, which all use the GNU C Library).
 

	
 
Unlike existing GNU application software, however, the licensing
 
implications of releasing the GNU C Library (``glibc'') under the GPL were
 
somewhat different. Applications released under the GPL would never
 
themselves become part of proprietary software. However, if glibc were
 
released under the GPL, it would require that any application distributed for
 
the GNU/Linux platform be released under the GPL\@.
 

	
 
Since all applications on a Unix-like system depend on the C library, it
 
means that they must link with that library to function on the system. In
 
other words, all applications running on a Unix-like system must be
 
combined with the C library to form a new whole derivative work that is
 
composed of the original application and the C library. Thus, if glibc
 
were GPL'd, each and every application distributed for use on GNU/Linux
 
would also need to be GPL'd, since to even function, such applications
 
would need to be combined into larger derivative works by linking with
 
glibc.
 

	
 
At first glance, such an outcome seems like a windfall for Free Software
 
advocates, since it stops all proprietary software development on
 
GNU/Linux systems. However, the outcome is a bit more subtle. In a world
 
where many C libraries already exist, many of which could easily be ported
 
to GNU/Linux, a GPL'd glibc would be unlikely to succeed. Proprietary
 
vendors would see the excellent opportunity to license their C libraries
 
to anyone who wished to write proprietary software for GNU/Linux systems.
 
The de-facto standard for the C library on GNU/Linux would likely be not
 
glibc, but the most popular proprietary one.
 

	
 
Meanwhile, the actual goal of releasing glibc under the GPL --- to ensure no
 
proprietary applications on GNU/Linux --- would be unattainable in this
 
scenario. Furthermore, users of those proprietary applications would also
 
be users of a proprietary C library, not the Free glibc.
 

	
 
The Lesser GPL was initially conceived to handle this scenario. It was
 
clear that the existence of proprietary applications for GNU/Linux was
 
inevitable. Since there were so many C libraries already in existence, a
 
new one under the GPL would not stop that tide. However, if the new C library
 
were released under a license that permitted proprietary applications
 
to link with it, but made sure that the library itself remained Free,
 
an ancillary goal could be met. Users of proprietary applications, while
 
they would not have the freedom to copy, share, modify and redistribute
 
the application itself, would have the freedom to do so with respect to
 
the C library.
 

	
 
There was no way the license of glibc could stop or even slow the creation
 
of proprietary applications on GNU/Linux. However, loosening the
 
restrictions on the licensing of glibc ensured that nearly all proprietary
 
applications at least used a Free C library rather than a proprietary one.
 
This trade-off is central to the reasoning behind the LGPL\@.
 

	
 
Of course, many people who use the LGPL today are not thinking in these
 
terms. In fact, they are often choosing the LGPL because they are looking
 
for a ``compromise'' between the GPL and the X11-style liberal licensing.
 
However, understanding FSF's reasoning behind the creation of the LGPL is
 
helpful when studying the license.
 

	
 

	
 
\section{What's the Same?}
 

	
 
Much of the text of the LGPL is identical to the GPL\@. As we begin our
 
discussion of the LGPL, we will first eliminate the sections that are
 
identical, or that have the minor modification changing the word
 
``Program'' to ``Library.''
 

	
 
First, LGPLv2.1~\S1, the rules for verbatim copying of source, are
 
equivalent to those in GPLv2~\S1.
 

	
 
Second, LGPLv2.1~\S8 is equivalent GPLv2~\S4\@. In both licenses, this
 
section handles termination in precisely the same manner.
 

	
 
LGPLv2.1~\S9 is equivalent to GPLv2~\S5\@. Both sections assert that
 
the license is a copyright license, and handle the acceptance of those
 
copyright terms.
 

	
 
LGPLv2.1~\S10 is equivalent to GPLv2~\S6. They both protect the
 
distribution system of Free Software under these licenses, to ensure that
 
up, down, and throughout the distribution chain, each recipient of the
 
software receives identical rights under the license and no other
 
restrictions are imposed.
 

	
 
LGPLv2.1~\S11 is GPLv2~\S7. As discussed, it is used to ensure that
 
other claims and legal realities, such as patent licenses and court
 
judgments, do not trump the rights and permissions granted by these
 
licenses, and requires that distribution be halted if such a trump is
 
known to exist.
 

	
 
LGPLv2.1~\S12 adds the same features as GPLv2~\S8. These sections are
 
used to allow original copyright holders to forbid distribution in
 
countries with draconian laws that would otherwise contradict these
 
licenses.
 

	
 
LGPLv2.1~\S13 sets up the FSF as the steward of the LGPL, just as GPLv2~\S9
 
does for GPL. Meanwhile, LGPLv2.1~\S14 reminds licensees that copyright
 
holders can grant exceptions to the terms of LGPL, just as GPLv2~\S10
 
reminds licensees of the same thing.
 

	
 
Finally, the assertions of no warranty and limitations of liability are
 
identical; thus LGPLv2.1~\S15 and LGPLv2.1~\S16 are the same as GPLv2~\S11 and \S
 
12.
 

	
 
As we see, the entire latter half of the license is identical.
 
The parts which set up the legal boundaries and meta-rules for the license
 
are the same. It is our intent that the two licenses operate under the
 
same legal mechanisms and are enforced precisely the same way.
 

	
 
We strike a difference only in the early portions of the license.
 
Namely, in the LGPL we go into deeper detail of granting various permissions to
 
create derivative works, so the re-distributors can make
 
some proprietary derivatives. Since we simply do not allow the
 
license to stretch as far as copyright law does regarding what
 
derivative works must be relicensed under the same terms, we must go
 
further to explain which derivative works we will allow to be
 
proprietary. Thus, we'll see that the front matter of the LGPL is a
 
bit more wordy and detailed with regards to the permissions granted to
 
those who modify or redistribute the software.
 

	
 
\section{Additions to the Preamble}
 

	
 
Most of the LGPL's Preamble is identical, but the last seven paragraphs
 
introduce the concepts and reasoning behind creation of the license,
 
presenting a more generalized and briefer version of the story with which
 
we began our consideration of the LGPL\@.
 

	
 
In short, FSF designed the LGPL for those edge cases where the freedom of the
 
public can better be served by a more lax licensing system. FSF doesn't
 
encourage use of the LGPL automatically for any software that happens to be a
 
library; rather, FSF suggests that it only be used in specific cases, such
 
as the following:
 

	
 
\begin{itemize}
 

	
 
\item To encourage the widest possible use of a Free Software library, so
 
  it becomes a de-facto standard over similar, although not
 
  interface-identical, proprietary alternatives
 

	
 
\item To encourage use of a Free Software library that already has
 
  interface-identical proprietary competitors that are more developed
 

	
 
\item To allow a greater number of users to get freedom, by encouraging
 
  proprietary companies to pick a Free alternative for its otherwise
 
  proprietary products
 

	
 
\end{itemize}
 

	
 
The LGPL's preamble sets forth the limits to which the license seeks to go in
 
chasing these goals. The LGPL is designed to ensure that users who happen to
 
acquire software linked with such libraries have full freedoms with
 
respect to that library. They should have the ability to upgrade to a newer
 
or modified Free version or to make their own modifications, even if they
 
cannot modify the primary software program that links to that library.
 

	
 
Finally, the preamble introduces two terms used throughout the license to
 
clarify between the different types of derivative works: ``works that use
 
the library,'' and ``works based on the library.''  Unlike the GPL, the LGPL must
 
draw some lines regarding derivative works. We do this here in this
 
license because we specifically seek to liberalize the rights afforded to
 
those who make derivative works. In the GPL, we reach as far as copyright law
 
allows. In the LGPL, we want to draw a line that allows some derivative works
 
copyright law would otherwise prohibit if the copyright holder exercised
 
his full permitted controls over the work.
 

	
 
\section{An Application: A Work that Uses the Library}
 

	
 
In the effort to allow certain proprietary derivative works and prohibit
 
others, the LGPL distinguishes between two classes of derivative works:
 
``works based on the library,'' and ``works that use the library.''  The
 
distinction is drawn on the bright line of binary (or runtime) derivative
 
works and source code derivatives. We will first consider the definition
 
of a ``work that uses the library,'' which is set forth in LGPLv2.1~\S5.
 

	
 
We noted in our discussion of GPLv2~\S3 (discussed in
 
Section~\ref{GPLv2s3} of this document) that binary programs when
 
compiled and linked with GPL'd software are derivative works of that GPL'd
 
software. This includes both linking that happens at compile-time (when
 
the binary is created) or at runtime (when the binary -- including library
 
and main program both -- is loaded into memory by the user). In GPL,
 
binary derivative works are controlled by the terms of the license (in GPLv2~\S3),
 
and distributors of such binary derivatives must release full
 
corresponding source\@.
 

	
 
In the case of LGPL, these are precisely the types of derivative works
 
we wish to permit. This scenario, defined in LGPL as ``a work that uses
 
the library,'' works as follows:
 

	
 
\newcommand{\workl}{$\mathcal{L}$}
 
\newcommand{\lplusi}{$\mathcal{L\!\!+\!\!I}$}
 

	
 
\begin{itemize}
 

	
 
\item A new copyright holder creates a separate and independent work,
 
  \worki{}, that makes interface calls (e.g., function calls) to the
 
  LGPL'd work, called \workl{}, whose copyright is held by some other
 
  party. Note that since \worki{} and \workl{} are separate and
 
  independent works, there is no copyright obligation on this new copyright
 
  holder with regard to the licensing of \worki{}, at least with regard to
 
  the source code.
 

	
 
\item The new copyright holder, for her software to be useful, realizes
 
  that it cannot run without combining \worki{} and \workl{}.
 
  Specifically, when she creates a running binary program, that running
 
  binary must be a derivative work, called \lplusi{}, that the user can
 
  run.
 

	
 
\item Since \lplusi{} is a derivative work of both \worki{} and \workl{},
 
  the license of \workl{} (the LGPL) can put restrictions on the license
 
  of \lplusi{}. In fact, this is what the LGPL does.
 

	
 
\end{itemize}
 

	
 
We will talk about the specific restrictions LGPLv2.1 places on ``works
 
that use the library'' in detail in Section~\ref{lgpl-section-6}. For
 
now, focus on the logic related to how the LGPLv2.1 places requirements on
 
the license of \lplusi{}. Note, first of all, the similarity between
 
this explanation and that in Section~\ref{separate-and-independent},
 
which discussed the combination of otherwise separate and independent
 
works with GPL'd code. Effectively, what LGPLv2.1 does is say that when a
 
new work is otherwise separate and independent, but has interface
 
calls out to an LGPL'd library, then it is considered a ``work that
 
uses the library.''
 

	
 
In addition, the only reason that LGPLv2.1 has any control over the licensing
 
of a ``work that uses the library'' is for the same reason that GPL has
 
some say over separate and independent works. Namely, such controls exist
 
because the {\em binary combination\/} (\lplusi{}) that must be created to
 
make the separate work (\worki{}) at all useful is a derivative work of
 
the LGPLv2.1'd software (\workl{}).
 

	
 
Thus, a two-question test that will help indicate if a particular work is
 
a ``work that uses the library'' under LGPLv2.1 is as follows:
 

	
 
\begin{enumerate}
 

	
 
\item Is the source code of the new copyrighted work, \worki{}, a
 
  completely independent work that stands by itself, and includes no
 
  source code from \workl{}?
 

	
 
\item When the source code is compiled, does it create a derivative work
 
  by combining with \workl{}, either by static (compile-time) or dynamic
 
  (runtime) linking, to create a new binary work, \lplusi{}?
 
\end{enumerate}
 

	
 
If the answers to both questions are ``yes,'' then \worki{} is most likely
 
a ``work that uses the library.''  If the answer to the first question
 
``yes,'' but the answer to the second question is ``no,'' then most likely
 
\worki{} is neither a ``work that uses the library'' nor a ``work based on
 
the library.''  If the answer to the first question is ``no,'' but the
 
answer to the second question is ``yes,'' then an investigation into
 
whether or not \worki{} is in fact a ``work based on the library'' is
 
warranted.
 

	
 
\section{The Library, and Works Based On It}
 

	
 
In short, a ``work based on the library'' could be defined as any
 
derivative work of LGPL'd software that cannot otherwise fit the
 
definition of a ``work that uses the library.''  A ``work based on the
 
library'' extends the full width and depth of copyright derivative works,
 
in the same sense that the GPL does.
 

	
 
Most typically, one creates a ``work based on the library'' by directly
 
modifying the source of the library. Such a work could also be created by
 
tightly integrating new software with the library. The lines are no doubt
 
fuzzy, just as they are with GPL'd works, since copyright law gives us no
 
litmus test for derivative works of a software program.
 

	
 
Thus, the test to use when considering whether something is a ``work
 
based on the library'' is as follows:
 

	
 
\begin{enumerate}
 

	
 
\item Is the new work, when in source form, a derivative work under
 
  copyright law of the LGPL'd work?
 

	
 
\item Is there no way in which the new work fits the definition of a
 
  ``work that uses the library''?
 
\end{enumerate}
 

	
 

	
 
If the answer is ``yes'' to both these questions, then you most likely
 
have a ``work based on the library.''  If the answer is ``no'' to the
 
first but ``yes'' to the second, you are in a gray area between ``work
 
based on the library'' and a ``work that uses the library.''
 

	
 
In our years of work with the LGPLv2.1, however, we have never seen a work
 
of software that was not clearly one or the other; the line is quite
 
bright. At times, though, we have seen cases where a derivative work
 
appeared in some ways to be a work that used the library and in other
 
ways a work based on the library. We overcame this problem by
 
dividing the work into smaller subunits. It was soon discovered that
 
what we actually had were three distinct components: the original
 
LGPL'd work, a specific set of works that used that library, and a
 
specific set of works that were based on the library. Once such
 
distinctions are established, the licensing for each component can be
 
considered independently and the LGPLv2.1 applied to each work as
 
prescribed.
 

	
 

	
 
\section{Subtleties in Defining the Application}
 

	
 
In our discussion of the definition of ``works that use the library,'' we
 
left out a few more complex details that relate to lower-level programming
 
details. The fourth paragraph of LGPLv2.1~\S5 covers these complexities,
 
and it has been a source of great confusion. Part of the confusion comes
 
because a deep understanding of how compiler programs work is nearly
 
mandatory to grasp the subtle nature of what LGPLv2.1~\S5, \P 4 seeks to
 
cover. It helps some to note that this is a border case that we cover in
 
the license only so that when such a border case is hit, the implications
 
of using the LGPL continue in the expected way.
 

	
 
To understand this subtle point, we must recall the way that a compiler
 
operates. The compiler first generates object code, which are the binary
 
representations of various programming modules. Each of those modules is
 
usually not useful by itself; it becomes useful to a user of a full program
 
when those modules are {\em linked\/} into a full binary executable.
 

	
 
As we have discussed, the assembly of modules can happen at compile-time
 
or at runtime. Legally, there is no distinction between the two --- both
 
create a derivative work by copying and combining portions of one work and
 
mixing them with another. However, under LGPL, there is a case in the
 
compilation process where the legal implications are different.
 
Specifically, while we know that a ``work that uses the library'' is one
 
whose final binary is a derivative work, but whose source is not, there
 
are cases where the object code --- that intermediate step between source
 
and final binary --- is a derivative work created by copying verbatim code
 
from the LGPL'd software.
 

	
 
For efficiency, when a compiler turns source code into object code, it
 
sometimes places literal portions of the copyrighted library code into the
 
object code for an otherwise separate independent work. In the normal
 
scenario, the derivative would not be created until final assembly and
 
linking of the executable occurred. However, when the compiler does this
 
efficiency optimization, at the intermediate object code step, a
 
derivative work is created.
 

	
 
LGPLv2.1~\S5\P4 is designed to handle this specific case. The intent of
 
the license is clearly that simply compiling software to ``make use'' of
 
the library does not in itself cause the compiled work to be a ``work
 
based on the library.''  However, since the compiler copies verbatim,
 
copyrighted portions of the library into the object code for the otherwise
 
separate and independent work, it would actually cause that object file to be a
 
``work based on the library.''  It is not FSF's intent that a mere
 
compilation idiosyncrasy would change the requirements on the users of the
 
LGPLv2.1'd software. This paragraph removes that restriction, allowing the
 
implications of the license to be the same regardless of the specific
 
mechanisms the compiler uses underneath to create the ``work that uses the
 
library.''
 

	
 
As it turns out, we have only once had anyone worry about this specific
 
idiosyncrasy, because that particular vendor wanted to ship object code
 
(rather than final binaries) to their customers and was worried about
 
this edge condition. The intent of clarifying this edge condition is
 
primarily to quell the worries of software engineers who understand the
 
level of verbatim code copying that a compiler often does, and to help
 
them understand that the full implications of LGPLv2.1 are the same regardless
 
of the details of the compilation progress.
 

	
 
\section{LGPLv2.1~\S6 \& LGPLv2.1~\S5: Combining the Works}
 
\label{lgpl-section-6}
 
Now that we have established a good working definition of works that
 
``use'' and works that ``are based on'' the library, we will consider the
 
rules for distributing these two different works.
 

	
 
The rules for distributing ``works that use the library'' are covered in
 
LGPLv2.1~\S6\@. LGPLv2.1~\S6 is much like GPLv2~\S3, as it requires the release
 
of source when a binary version of the LGPL'd software is released. Of
 
course, it only requires that source code for the library itself be made
 
available. The work that ``uses'' the library need not be provided in
 
source form. However, there are also conditions in LGPLv2.1~\S6 to make sure
 
that a user who wishes to modify or update the library can do so.
 

	
 
LGPLv2.1~\S6 lists five choices with regard to supplying library source
 
and granting the freedom to modify that library source to users. We
 
will first consider the option given by \S~6(b), which describes the
 
most common way currently used for LGPLv2.1 compliance on a ``work that
 
uses the library.''
 

	
 
LGPLv2.1~\S6(b) allows the distributor of a ``work that uses the library'' to
 
simply use a dynamically linked, shared library mechanism to link with the
 
library. This is by far the easiest and most straightforward option for
 
distribution. In this case, the executable of the work that uses the
 
library will contain only the ``stub code'' that is put in place by the
 
shared library mechanism, and at runtime the executable will combine with
 
the shared version of the library already resident on the user's computer.
 
If such a mechanism is used, it must allow the user to upgrade and
 
replace the library with interface-compatible versions and still be able
 
to use the ``work that uses the library.''  However, all modern shared
 
library mechanisms function as such, and thus LGPLv2.1~\S6(b) is the simplest
 
option, since it does not even require that the distributor of the ``work 
 
based on the library'' ship copies of the library itself.
 

	
 
LGPLv2.1~\S6(a) is the option to use when, for some reason, a shared library
 
mechanism cannot be used. It requires that the source for the library be
 
included, in the typical GPL fashion, but it also has a requirement beyond
 
that. The user must be able to exercise her freedom to modify the library
 
to its fullest extent, and that means recombining it with the ``work based
 
on the library.''  If the full binary is linked without a shared library
 
mechanism, the user must have available the object code for the ``work
 
based on the library,'' so that the user can relink the application and
 
build a new binary.
 

	
 
The remaining options in LGPLv2.1~\S6 are very similar to the other choices
 
provided by GPLv2~\S3. There are some additional options, but time does
 
not permit us in this course to go into those additional options. In
 
almost all cases of distribution under LGPL, either LGPLv2.1~\S6(a) or LGPLv2.1~\S6(b) are
 
exercised.
 

	
 
\section{Distribution of the Combined Works}
 

	
 
Essentially, ``works based on the library'' must be distributed under the
 
same conditions as works under full GPL\@. In fact, we note that 
 
LGPLv2.1~\S2 is nearly identical in its terms and requirements to GPLv2~\S2.
 
There are again subtle differences and additions, which time does not
 
permit us to cover in this course.
 

	
 
\section{And the Rest}
 

	
 
The remaining variations between the LGPL and the GPL cover the following
 
conditions:
 

	
 
\begin{itemize}
 

	
 
\item Allowing a licensing ``upgrade'' from the LGPL to the GPL\@ (in LGPLv2.1~\S3)
 

	
 
\item Binary distribution of the library only, covered in LGPLv2.1~\S4,
 
  which is effectively equivalent to LGPLv2.1~\S3
 

	
 
\item Creating aggregates of libraries that are not derivative works of
 
  each other, and distributing them as a unit (in LGPLv2.1~\S7)
 

	
 
\end{itemize}
 

	
 

	
 
Due to time constraints, we cannot cover these additional terms in detail,
 
but they are mostly straightforward. The key to understanding LGPLv2.1 is
 
understanding the difference between a ``work based on the library'' and a
 
``work that uses the library.''  Once that distinction is clear, the
 
remainder of LGPLv2.1 is close enough to GPL that the concepts discussed in
 
our more extensive GPL unit can be directly applied.
 

	
 
%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
 
\chapter{Integrating the GPL into Business Practices}
 

	
 
Since GPL'd software is now extremely prevalent through the industry, it
 
is useful to have some basic knowledge about using GPL'd software in
 
business and how to build business models around GPL'd software.
 

	
 
\section{Using GPL'd Software In-House}
 

	
 
As discussed in Sections~\ref{GPLv2s0} and~\ref{GPLv2s5} of this tutorial,
 
the GPL only governs the activities of copying, modifying and
 
distributing software programs that are not governed by the license.
 
Thus, in FSF's view, simply installing the software on a machine and
 
using it is not controlled or limited in any way by the GPL\@. Using Free
 
Software in general requires substantially fewer agreements and less
 
license compliance activity than any known proprietary software.
 

	
 
Even if a company engages heavily in copying the software throughout the
 
enterprise, such copying is not only permitted by GPLv2~\S\S1 and 3, but it is
 
encouraged!  If the company simply deploys unmodified (or even modified)
 
Free Software throughout the organization for its employees to use, the
 
obligations under the license are very minimal. Using Free Software has a
 
substantially lower cost of ownership --- both in licensing fees and in
 
licensing checking and handling -- than the proprietary software
 
equivalents.
 

	
 
\section{Business Models}
 
\label{Business Models}
 

	
 
Using Free Software in house is certainly helpful, but a thriving
 
market for Free Software-oriented business models also exists. There is the
 
traditional model of selling copies of Free Software distributions.
 
Many companies make substantial revenue
 
from this model. Some choose this model because they have
 
found that for higher-end hardware, the cost of the profit made from
 
proprietary software licensing fees is negligible. The real profit is
 
in the hardware, but it is essential that software be stable, reliable
 
and dependable, and the users be allowed to have unfettered access to
 
it. Free Software, and GPL'd software in particular (because IBM can
 
be assured that proprietary versions of the same software will not
 
exist to compete on their hardware) is the right choice.
 

	
 
For example, charging a ``convenience fee'' for Free Software,
 
when set at a reasonable price (around \$60 or so), can produce some
 
profit. Even though Red Hat's system is fully downloadable on their
 
Web site, people still go to local computer stores and buy copies of their
 
box set, which is simply a printed version of the manual (available under
 
a Free license as well) and the Free Software system it documents.
 

	
 
\medskip
 

	
 
However, custom support, service, and software improvement contracts
 
are the most widely used models for GPL'd software. The GPL is
 
central to their success, because it ensures that the code base
 
remains common, and that large and small companies are on equal
 
footing for access to the technology. Consider, for example, the GNU
 
Compiler Collection (GCC). Cygnus Solutions, a company started in the
 
early 1990s, was able to grow steadily simply by providing services
 
for GCC --- mostly consisting of new ports of GCC to different or new,
 
embedded targets. Eventually, Cygnus was so successful that
 
it was purchased by Red Hat where it remains a profitable division.
 

	
 
However, there are very small companies that compete in
 
this space. Because the code-base is protect by the GPL, it creates and
 
demands industry trust. Companies can cooperate on the software and
 
improve it for everyone. Meanwhile, companies who rely on GCC for their
 
work are happy to pay for improvements, and for ports to new target
 
platforms. Nearly all the changes fold back into the standard
 
versions, and those forks that exist remain freely available.
 

	
 
\medskip
 

	
 
\label{Proprietary Relicensing}
 

	
 
A final common business model that is perhaps the most controversial is
 
proprietary relicensing of a GPL'd code base. This is only an option for
 
software in which a particular entity is the sole copyright holder. As
 
discussed earlier in this tutorial, a copyright holder is permitted under
 
copyright law to license a software system under her copyright as many
 
different ways as she likes to as many different parties as she wishes.
 

	
 
Some companies use this to their
 
financial advantage with regard to a GPL'd code base. The standard
 
version is available from the company under the terms of the GPL\@.
 
However, parties can purchase separate proprietary software licensing for
 
a fee.
 

	
 
This business model is at best problematic and at worst predatory because it means that the GPL'd code
 
base must be developed in a somewhat monolithic way, because volunteer
 
Free Software developers may be reluctant to assign their copyrights to
 
the company because it will not promise to always and forever license the
 
software as Free Software. Indeed, the company will surely use such code
 
contributions in proprietary versions licensed for fees.
 

	
 
\section{Ongoing Compliance}
 

	
 
GPL compliance is in fact a very simple matter --- much simpler than
 
typical proprietary software agreements and EULAs. Usually, the most
 
difficult hurdle is changing from a proprietary software mindset to one
 
that seeks to foster a community of sharing and mutual support. Certainly
 
complying with the GPL from a users' perspective gives substantially fewer
 
headaches than proprietary license compliance.
 

	
 
For those who go into the business of distributing {\em modified}
 
versions of GPL'd software, the burden is a bit higher, but not by
 
much. The glib answer is that by releasing the whole product as Free
 
Software, it is always easy to comply with the GPL. However,
 
admittedly to the dismay of FSF, many modern and complex software
 
systems are built using both proprietary and GPL'd components that are
 
not legally derivative works of each other. Sometimes, it is easier simply to
 
improve existing GPL'd application than to start from scratch. In
 
exchange for that benefit, the license requires that the modifier give
 
back to the commons that made the work easier in the first place. It is a
 
reasonable trade-off and a way to help build a better world while also
 
making a profit.
 

	
 
Note that FSF does provide services to assist companies who need
 
assistance in complying with the GPL. You can contact FSF's GPL
 
Compliance Labs at $<$licensing@fsf.org$>$.
 

	
 
%FIXME-LATER: should have \tutorialpart
 

	
 
If you are particularly interested in matters of GPL compliance, we
 
recommend the next two parts, which include both recommendations on good
 
compliance and compliance case studies.
 

	
 
% =====================================================================
 
% END OF FIRST DAY SEMINAR SECTION
 
% =====================================================================
 

	
 
%%  LocalWords:  Sebro Novalis Ravicher GPLv GPL'd copylefted LGPLv OSI USC
 
%%  LocalWords:  noncommercially counterintuitive Berne copyrightable DRM UC
 
%%  LocalWords:  proprietarize proprietarization Stallman's Tridgell's RMS
 
%%  LocalWords:  Lessig Lessig's Stallman Proto GPLs proto Tai pre GPL's ful
 
%%  LocalWords:  legalbol AGPLv Runtime licensor licensors relicense UCITA
 
%%  LocalWords:  unprotectable Intl nd th Kepner Tregoe Bando Indust Mitel
 
%%  LocalWords:  Iqtel Bateman Mitek Arce protectable hoc faire de minimis
 
%%  LocalWords:  Borland Int'l uncopyrightable LLC APIs Ent Connectix DVD's
 
%%  LocalWords:  redistributor diachronic unshared subpart redistributors
 
%%  LocalWords:  CDs userbase reshifts licensor's distributee impliedly Mgmt
 
%%  LocalWords:  patentee  relicenses irrevocability Jacobsen Katzer TRW CCS
 
%%  LocalWords:  Unfreedonia administrivia Relicensing impermissibly centric
 
%%  LocalWords:  permissibility firehose bytecode minified Javascript DLLs
 
%%  LocalWords:  preprocessors functionalities offsite sublicensing DMCA CFR
 
%%  LocalWords:  anticircumvention WIPO BitTorrent multidirectional Magnuson
 
%%  LocalWords:  subdefinition Dryvit Stroebner Tandy TRS superset LGPL SLES
 
%%  LocalWords:  cryptographic relicensing removability sublicensed Novell
 
%%  LocalWords:  anticompetitive administrability sublicensable licensable
 
%%  LocalWords:  sublicense sublicensees sublicenses affixation Novell's
 
%%  LocalWords:  severability Affero LGPL'd lingua franca glibc facto LGPL's
 
%%  LocalWords:  relicensed runtime subunits relink downloadable MontaVista
 
%%  LocalWords:  CodeSourcery OpenTV MySQL TrollTech
0 comments (0 inline, 0 general)