If you've come in in the middle, you may want to read the <-- Intro
Typical analogies for copyright infringement break down when confronted with this situation. This is because the lawsuit has an unusual structure as much as because the accused work is atypical.
The usual plaintiff claiming infringement of computer software targets one of four kinds of defendants: n1
a former employee who's now started a new business that competes with the plaintiff, or
a third party who has infringed the plaintiff's work and now seeks commercial advantage from it, or
a straightforward counterfeiter, or
a licensee of the work who is operating outside the bounds of the license.
Here the defendant (an end user) is not a former employee
of the plaintiff. They've not set up shop in competition with SCO.
There is no direct commercial advantage to the individual end user,
as they are not selling the code. n2
The end user is not relabeling SCO's work as their own and marketing
it - nor is their use of the code akin to a "repurposing"
of electronic information for business gain. n3
This isn't a simple case of CD Pirates having stamped out (well,
injected
out, actually) disc after disc of SCO's product.
Finally, there's no license at all between the end user and
SCO.
SCO must reach through all the links in the chain
to target the final user. In the traditional copyright areas, this
sort of suit is not even possible. The ultimate possessor
of a copy of a creative work can't be sued for using it.
Consider the analogous situation in the literary context:
An editor assembles an anthology of poetry, with some of her own work but also including the submissions of a score of other authors. The anthology is published and distributed in book form, and an "end user" buys a copy. If someone then pops up to claim that their work has been infringed, that in fact whole sections of their poems appear in the anthology without their permission, the end user has no legal liability. The editor, publisher, distributor, and vendor may be legally liable - but the end user is not. The end user is free to read, stain, mark, burn, or prop up a table leg with the book. None of those normal uses infringe the rights of the claimant. Unless the end user:
copies the poem in question, or
sells such copies, or
writes another poem that draws very heavily on one of the originals, or
reads the poem aloud in front of a paying crowd, or
posts it on a billboard in Times Square or another public place,
they're not in violation of the Copyright Act n4. All the wrong to the plaintiff was done before the end user took the book home. Without a copyright violation, the claimant can't win.
The key difference in the software context is that the end
user of the Linux kernel must copy it to a computer's hard drive to
install it. The user also "copies"
the kernel when the computer boots and the object code is loaded into
RAM to run. In the case of the Linux kernel user, there are
ongoing violations of the claimant's rights.
The Linux kernel source tree is an unusual piece of software. Unlike the typical proprietary computer program, which is a "work made for hire" owned by the company whose employees created it, Linux is a collective work. It consists of or contains roughly:
8669 Copyright notices by approx. 1100 distinct authors found in 10109 source files containing 4700792 lines of code total. n5
SCO has not specifically claimed that the kernel as a whole is a derivative work of UNIX. Only parts of the kernel may be infringing. According to SCO, it is particularly the "enterprise ready" features of UNIX that have been copied into Linux. But Linux runs on at least thirty different hardware platforms, and much of the code in the kernel is dependent on the particular hardware platform. Only a relatively few source code files are in every Linux kernel. A laptop is probably not equipped with the hardware required to support many "enterprise ready" features. , so a kernel that works on the laptop may be built without the And that leads to another metaphor:
An artist writes a symphony, a musical
composition, with parts for all the sections of a full orchestra.
After the symphony is published, copies of the composition are made
available and license is granted for anyone to play it. A small
chamber orchestra gets a copy, and extracts only the parts of the
symphony that are appropriate for their instruments. For instance,
say the chamber orchestra only includes strings - they only use the
parts for stringed instruments. Assume they play the (partial)
symphony.
Suppose another composer appears after the symphony
is published and alleges that certain sections of the full
symphony are literal copies of that composer's work. Say that the
claimed sections are for brass instruments, does our chamber
orchestra infringe? That is, granting for the sake of argument that a
full orchestra playing the entire composition would infringe the
rights of the second composer, can we say that the chamber
orchestra's performance does so?
So far as the author of this analysis is able to determine, there's no court decision establishing the answer to that question. Laymen tend to instantly answer it in the negative, so SCO's claim against an end user may depend on which "enterprise ready" features the defendant is using.
Even with the unusual structure of the suit and the relatively odd accused work, it's still just a copyright infringement suit. A plaintiff claiming copyright violation must show two things, 1) Ownership of a valid copyright, and 2) Infringement of an exclusive right. n5.5
Ownership is prima facie established by copyright registration, and SCO did register its claim to UNIX at the Copyright Office in July of 2003. The Registration Record lists that a "printout only" was deposited; it's not clear from the Registration Record whether SCO deposited all of the UNIX source code, or only the partial deposit allowed by the statute. If the latter, SCO may not be entitled to the presumption, since at least one Federal District Court has held that a partial deposit of a claimed computer program failed to identify the work claimed. n6
If SCO can not rely on its registration alone, SCO must produce other evidence of copyright ownership to get started. The defendant would then have the challenge of countering such evidence with proof of either the non-copyrightability of the code, or proof that someone else is the owner of copyright in the code. It's a bit too simplistic to say, "the code", since UNIX is made up of thousands of individual files. A robust claim of infringement would have to name particular files or sections of files (broadly, "works") that were copied into Linux. Copyrightability or ownership of each of those particular works may individually be disputed.
The Merger doctrine - Section
102(b) of the Copyright Act specifies that copyright does not
protect "any idea, procedure, process, system, method of
operation, concept, principle, or discovery". An underlying
idea is not copyrightable - only a particular expression of
an idea is. The Merger doctrine says that if an idea admits of only
a very few possible forms of expression, the idea "merges"
with its expression; and the merged idea-expression is not
protectable by Copyright.
Computer programs must be very
precise in order to cause correct operation of the underlying
machine. That requirement for precision may indicate that the idea
behind the program must be expressed in a particular way; if
so, that expression is not copyrightable. n7
Features dictated by outside forces or interoperability
- UNIX and Linux, as well as many licensed UNIX variants, conform to
a number of U.S. and International standards that define interfaces
for computer environments. n8
These standards dictate, for instance, the type and number of
arguments given to kernel function calls, as well as the names of
the function calls themselves.
A program that implements
such a standard interface also works within defined external
constraints that will limit the programmer's choices of efficient
internal structures. The aspects of the Linux kernel which are
dictated by those standards can not deviate from the same aspects of
UNIX - to do so would be to violate the standards and result in a
program whose utility was severely reduced.
The scenes a fair doctrine - another constraint on rampant creativity has to do with good programming practices. One of the authors of the C programming language (both UNIX and Linux are written in C) has urged programmers to "use canonical forms". n9 This admonishment and the underlying benefits of that good practice leads to control structures in different C programs looking very, very similar to one another. Under the scenes a fair doctrine, copyright protection may be precluded where the work contains features which are indispensable or standard for such a work.
Too brief - just too little expression to meet the standard of creativity required by the Copyright Act.
Since what SCO is asserting is that "enterprise ready" elements of UNIX are at the heart of its heartburn, the "too brief" objection may not have play here. That's because the critical code for the sophisticated features SCO has named is complex and lengthy. The other objections may knock some sections of UNIX code out from under Copyright protection. Presuming that some claimed sections are copyrightable, there is still a question of ownership of the copyright.
As established by an earlier suit
brought by Unix System Laboratories against the University of
California, some portions of UNIX source code are in the pubic domain
because of a technical lapse
possible under the previous version of the Copyright Act. Any work
which is a derivative of a work in the public domain may have its own
copyright in new or additional elements, but the derivative work does
not take the original out of the public domain. Both later
versions of UNIX and Linux may lawfully make use of those public
domain sections of UNIX, and neither UNIX nor Linux infringe the
other in their use of those sections.
Other sections of UNIX
are actually copied or derived from other operating systems. Those
sections, if copyrightable, are owned by their authors and not by
SCO. A recent section of source code shown by SCO, comprising a
function called the "Berkeley packet filter", is an example
of this. The Berkeley packet filter was designed and implemented
first by the operating system programmers at the University of
California and released as part of the "Berkeley Standard
Distribution". Both UNIX and Linux include nearly identical
sections of this source code, but its presence in both of them is not
due to copying from UNIX to Linux, but rather due to copying from BSD
into both UNIX and Linux.
SCO was and is a distributor of the Linux kernel. n12 As part of the free software model, SCO made Linux kernel source code available to everyone on its web site. Can they now turn around and sue the people that accepted their offer?
Notwithstanding the above defenses, there may be sections of UNIX
that SCO has valid copyright in. For those sections, SCO may dictate
how that section is reproduced, distributed, adapted, or displayed.
If an entity is taking any of the above actions without SCO's say-so,
that action infringes SCO's exclusive rights. n12.5
This presentation assumes that end users are the target of
SCO's claim. The kinds of infringing actions that might exist are
dependent on what kind of end user is targeted. Here's a
substantially complete taxonomy of Linux kernel end user types.
Object-code-only users, who make no derivative works and are unlikely to redistribute the object code;
Source-and-object-code users who either modify the manifest and create a new object code version of the kernel OR modify the source code files and then create a new object code version of the kernel;
Source-code-only users who are working with or on another program and want the source for reference or inclusion in that program;
Kernel-tree contributors who are refining, updating, and correcting the source files after which the patches will be submitted for inclusion in the next release of the tree.
The example defendants in this presentation fall into the "Object-code-only" end user camp, and as such, are most likely to infringe only the exclusive right of reproduction. The full explication of the other end user types and their possibly infringing actions is left as an exercise for the reader.
Technically, every objection raised above is a "defense"; but from this point on, it is assumed that those defenses have been countered and the plaintiff's case established. That is, despite the objections, SCO has shown their valid copyright in some portion of the Linux kernel. Once that point is reached, it's up to the end users to affirmatively justify their use of the kernel.
The end user denoted "Some Company" in the linked graphic is a commercial entity who paid for RHAS. This presentation assumes that Some Company's server makes use of source code that SCO can prove it owns copyright in, so Some Company's use of the Linux kernel does infringe SCO's copyright. Nevertheless, the commercial defendant has two affirmative defenses to liability for that infringement.
As has already been noted, there is no contract between the
defendant and the plaintiff - but there has been a contract formed
between the defendant and the vendor of the Linux distribution at
issue here. "Some Company" paid more than two thousand
dollars to a reputable seller of goods and services, for delivery of
a tangible product that the purchaser could reasonably expect to be
free of encumbrances. That the vendor of the product may have a
liability to a third party for the materials in the purchased product
is not the concern of "Some Company", so long as they acted
in good faith and had no notice of the third party claim.
That
statement, in a nutshell, represents an ancient legal concept that
goes by various names: In the real property and sales of chattels
contexts, such a person is called a "bona fide purchaser for
value"; in the common law of negotiable securities, the company
would be called a "holder in due course"; under the Uniform
Commercial Code the term is "protected purchaser". n13
All of these terms apply to an entity who's come into possession of a
piece of property in the normal course of commercial dealings, having
given something of value for the property, without reason to suspect
that the seller is without authority to dispose of the property. Such
an entity will be entitled to keep the property, even if the original
owner can positively prove that the vendor acted without authority.
The rationale for this result is based in the marketplace's need for
certainty: commercial actors need to be able to rely on proper
appearances and proper behaviour of sellers to establish an
unconstricted economic environment. n14
The
concept encourages due diligence on the part of original owners. If
they do not actively monitor their agents (or others who may have
access to their possessions) they are subject to loss.
The
defendant Some Company's purchase was not in a suspicious or shady
setting. RedHat's web store has been open for several years, and it
is accessible to anyone on the Internet. Some Company's
certainly gave value - many other Linux distributions are available
at lesser cost, including one from SCO itself. As to good faith and
notice, before the SCO v. IBM suit there was no reason for Some
Company to suspect that RedHat lacked authority to sell source code
or object code versions of the Linux kernel. The industry as a whole
was aware that Linux is available under the GPL, and that RedHat's
packaging and distribution of the Linux kernel was appropriate under
the terms of that license.
SCO might assert that the IBM suit
and SCO's press releases in mid-2003 gave notice to prospective
purchasers, and therefore that Some Company's purchase of RHAS after
midyear was either in bad faith or with notice of SCO's
claims.
SCO's own actions speak against that interpretation,
however. In the first place, SCO has not divulged the particular
files or code that it is claiming were infringed despite IBM's
discovery requests and calls on SCO by Linux kernel developers to do
so. SCO also continued to allow no-cost downloads of its own Linux
distribution on its own web site well into October of 2003. SCO's
threats against UNIX licensees and statements directed at end users
have been non-specific, and their announced intentions to pursue
enforcement actions have been rescheduled, postponed indefinitely,
canceled, or allowed to slip into obscurity. SCO's actions and
pronouncements lack the substance required to constitute effective
notice of a plausible right on their part. The rest of the computer
industry and the commercial world can not stop in place because one
member of the industry wishes to make unsubstantiated claims against
unnamed parties.
The power to give effective notice is
entirely SCO's. If they wish to defeat future claims of protected
purchasers, they need only produce at least one instance of plausible
infringement of UNIX. Until that is forthcoming, commercial
purchasers of Linux distributions are acting within reasonable
bounds.
The Copyright statute specifically recognizes that certain infringing uses of copyrighted works are nonetheless justified uses, and it exempts such uses from liability. Section 107 of the statute lays out four factors which are to be evaluated against the proposed use to determine if the use is "fair use":
the purpose and character of the use, including whether such use is of a commercial nature or is for nonprofit educational purposes;
the nature of the copyrighted work;
the amount and substantiality of the portion used in relation to the copyrighted work as a whole; and
the effect of the use upon the potential market for or value of the copyrighted work.
Courts' applications of the fair use factors have been very
fact-specific. In order to avoid generating a set of assumptions
about Some Company's use of the Linux kernel just to apply a flexible
doctrine to those assumptions, no more will be said about the
particular shape of this defense. An eventual defendant will want to
marshal all the beneficial facts at their disposal to strengthen
their claim of fair use.
The end user shown as "Some home user" in the linked graphic is an individual who downloaded rh9 and installed it on their PC without paying anything for the distribution. This situation brings out a curious aspect to SCO's claims, an aspect that may relieve the typical home user of any direct infringement at all.
Among home users, there may be a subset who make no use at all of
the ate_utils.c file. That's because the Linux kernel can be
complied for more than two dozen different computer platforms. Some
end users have x86 machines; some have PPC machines; some have SPARC
machines, and so on. The ate_utils.c file, however, is specific to
just one platform.
Even after SCO makes known the claimed
file names, there may be end users that do not make use of any of the
files. Since SCO claims that copying UNIX code aided in the
implementation of particular high-end features such as "Symmetric
Multi-Processing" or "Non-Uniform Memory Access", any
platform that doesn't use those features may not need any of the
claimed files.
If an object code kernel was built for a
platform that does not include any of those features, it may not
contain source code from files that infringe SCO's copyrights. Those
omitted files would be strictly unnecessary for the build. Their
contents would not have been consulted nor used to build the object
code kernel. Their presence in the kernel tree is not required for
that particular object code kernel to be produced. In short, there
would be no copy of UNIX code in such a kernel at all -- and so no
use of the claimed code -- and so no infringement.
If, however, SCO shows that even typical home users are running copyrighted code, this defendant may raise the defense of "fair use" just as the commercial user did. In general, a non-commercial user has a better chance of prevailing on the four factors commonly applied to evaluate fair use; but as above this presentation will make no attempt to invent particular detailed fact patterns solely to apply the test.
To complete this section, one more obscure approach SCO could
conceivably take is presented.
If no infringing file was
copied or transformed to build the object code kernel, that kernel
does not infringe either, unless the kernel source tree, once
assembled, is viewed as a single work. If the kernel tree is a
single work, who is its owner? One might be able to call the kernel
tree a joint work. A joint work is co-owned by every
contributor.
In order for SCO to sue an object-code-only user
whose object code kernel does not make use of any infringing file,
SCO would have to maintain that the kernel source tree is a single
work, and that use of any part of the source tree is infringement.
Then, even though a home user does not make direct use of any UNIX
code, that user still infringes the kernel source tree - which every
contributor owns as a joint work. There are (at least) two problems
with that approach:
Joint authors must have the intent to create a joint work: but SCO did not intend for the UNIX code to even be in the tree. They, the owners of copyright, can't have intended to produce a joint work if they did not even intend to contribute.
A joint work may be licensed in its entirety by any of its authors, and approximately all of the other authors would instantly license the work to everyone, thus depriving SCO of a claim.