The Copyright That Wasn't?

Mon, 13 Oct 2025

I was thinking about the ghosts of UNIX the other day, not in a supernatural sense, but wondering who actually holds the copyright on ancient UNIX? It's a question that seems straightforward, but asking it is like pulling a loose thread on a historical tapestry. Pull it, and you don't get a neat answer; you unravel a story of academic collaboration, corporate warfare, and two lawsuits that, when examined together, create a legal paradox so profound it calls into question the very existence of the copyright being fought over. This isn't a story about code; it's a detective story about a legal ghost - a copyright that has been bought, sold, and litigated, all while hiding a secret: it may not have been there at all.

The first half of the story begins not in a courtroom or a corporate boardroom, but at the University of California, Berkeley.

In the early 1970s, AT&T, where UNIX started, was operating under a consent decree that limited its ability to enter the computer business. As a result, its Bell Labs research division licensed early versions of UNIX to universities and research institutions for a nominal fee, including complete source code, although it wasn't free software.

At UC Berkeley, a team in the Computer Systems Research Group (CSRG) took the UNIX source code and began dramatically modifying and extending it. Beginning with Version 6 Unix, people like Bill Joy made what would become the Berkeley Software Distribution (BSD). The first release, 1BSD in 1978, wasn't a standalone operating system but a collection of add-ons and improvements to AT&T's V6.

By the late 1970s, the original home of UNIX - the 16-bit DEC PDP-11 - was showing its age. A critical moment came with the arrival of the 32-bit DEC VAX computer. In 1979, Bell Labs programmers Tom London and John F. Reiser undertook the task of porting the latest version of Research Unix, the Seventh Edition (V7), to the VAX architecture, and named it UNIX/32V. The Berkeley team, however, found 32V lacking, particularly in its failure to utilize the VAX's virtual memory capabilities. In a feat of engineering, they rewrote large parts of the 32V kernel to add this feature, releasing the result as 3BSD in 1979. This made UNIX/32V the "common ancestor" of the two great branches of the UNIX family tree: AT&T's own System V and Berkeley's 4BSD. It also placed 32V directly in the crosshairs of a future legal battle.

Over the next decade, the CSRG continued its work, funded in part by DARPA, to make a version of BSD that was completely free of proprietary AT&T code. This effort culminated in the 1991 release of Networking Release 2 (Net/2), a nearly complete operating system that Berkeley believed was "unencumbered" by AT&T code.

This release attracted commercial interest. A new company, Berkeley Software Design, Inc. (BSDi), took the Net/2 source code, filled in the few missing pieces, and began selling a version called BSD/386. This was a red flag for AT&T's subsidiary, Unix System Laboratories (USL), which held the rights to UNIX. In 1992, USL filed a lawsuit against BSDi focused on two main issues:

Trademark Infringement: USL objected to BSDi's use of the phone number "1-800-ITS-UNIX," arguing it was an unauthorized use of their registered trademark.
False Advertising: USL claimed that BSDi's promotional materials, which stated that BSD/386 didn't require a license from AT&T/USL, were false and misleading.

Notably, the original complaint didn't actually include a claim for copyright infringement. It only "reserved the right" to add such claims later.

BSDi's legal team immediately seized on this omission. In a motion to dismiss filed in May 1992, they pointed out the central contradiction in USL's case: the only way BSDi's advertising could be "false" was if BSD/386 contained proprietary UNIX code, which would constitute copyright infringement. Yet, USL hadn't actually sued them for infringement. BSDi's lawyers essentially called USL's bluff, arguing that USL couldn't sustain a false advertising claim without first proving the underlying infringement - something they suggested USL had no "good faith basis" to do at the time.

This strategic jab set the stage for a pivotal court hearing. At the hearing, BSDi's lawyers contended that BSD/386 was composed almost entirely of the Net/2 source code, which the University of California freely distributed. They had only added a few of their own files to complete the system. BSDi accepted liability for its own files but argued that it couldn't be held responsible for the thousands of other files it had obtained from the University.

The judge agreed with BSDi's reasoning and presented USL with an ultimatum: either restate the complaint to focus solely on the files BSDi had added, or he would dismiss the case entirely.

This put USL in an impossible position. Faced with the choice of fighting a tiny, ineffective legal battle or having their case thrown out, USL chose a third, far more aggressive option: Rather than narrowing their claim, they dramatically expanded it. In July 1992, USL filed the amended complaint. This new complaint not only targeted BSDi but also added the Regents of the University of California as a primary defendant. USL's claims were sweeping: copyright infringement, trade secret misappropriation, and trademark dilution. By going after the source of the code, USL escalated the fight into a full-blown copyright and trade secret lawsuit that would define the "UNIX Wars". USL sought a preliminary injunction to halt all distribution of Net/2 and BSD/386, arguing that they were illegal derivatives of proprietary UNIX code. The amended complaint was a direct consequence of BSDi's legal team successfully cornering them, forcing them to either put up or shut up on the core issue of copyright infringement.

For nearly a year, Berkeley and BSDi were on the defensive, facing a legal onslaught from a corporate giant. Then, in June 1993, just a few months after a crucial court ruling went in their favor, the University of California filed a countersuit against USL in California state court.

Berkeley's lawyers alleged that USL's own UNIX System V was itself a derivative work of BSD. They claimed that USL had incorporated massive amounts of code developed at Berkeley into System V without providing the attribution and credit required by the very license agreements that USL held with the University. The University's complaint was a mirror image of USL's, using the same principles of copyright and contractual obligation against USL.

Berkeley demanded that USL be forced to reprint all of its documentation to include the proper credits, run corrective advertisements in major business publications like The Wall Street Journal, and notify its entire licensee base of the oversight. This wasn't merely a defensive denial; it created a situation of mutually assured destruction: if USL were to win its case and establish a high standard for proving code provenance and enforcing license terms, that same high standard would immediately be turned against itself in the California countersuit. This dramatically increased the risk and cost of continued litigation for USL.

The case landed in the courtroom of New Jersey District Judge Dickinson R. Debevoise, who was tasked with deciding whether to grant USL's injunction. To do so, he had to assess whether USL was likely to succeed on the merits of its case. This required him to examine the validity of USL's copyright claim.

Under the U.S. copyright law in effect at the time, copyright wasn't automatic. To have a copyright, you were required to publish the work with a proper copyright notice affixed. The law was unforgiving: a "general publication" of a work without the necessary notice placed it directly and irrevocably into the public domain. This is a legal concept known as copyright forfeiture. BSDi's legal team argued that this was precisely what had happened with UNIX/32V. AT&T had released it in 1979, distributing thousands of copies to licensees without the mandatory copyright notice. This, they contended, was a general publication that forfeited any copyright AT&T might have had.

In any major litigation, the hearing on a motion for a preliminary injunction is a moment of high drama. It's often the first time the parties present the core of their arguments to the judge, and the court's decision, while not final, provides a powerful signal of how it views the case's underlying merits. For USL, the injunction it sought against Berkeley and BSDi was critical; it would have halted all distribution of the allegedly infringing software, effectively crippling its new competitor, BSDi, while the lengthy legal process unfolded. The hearing, held in late 1992, and the subsequent ruling issued by Judge Dickinson R. Debevoise of the U.S. District Court for the District of New Jersey in March 1993, proved pivotal in the entire lawsuit.

USL's argument for the injunction rested on the claim that it would suffer "irreparable harm" from the continued distribution of Net/2, which it alleged contained its copyrighted code and trade secrets. To win such an injunction, a plaintiff must demonstrate, among other things, a strong likelihood that they will ultimately win the case at trial. It was on this crucial point that USL's case crumbled.

Judge Debevoise didn't grant the injunction. Instead, he issued a detailed, 40-page opinion that amounted to a systematic dismantling of USL's position. He expressed profound skepticism about the validity of USL's copyright claims, particularly its copyright claim on UNIX/32V. This wasn't a ruling on a minor procedural point; it was a direct commentary on the very heart of USL's case.

The judge's reasoning was clear, direct, and devastating for USL. He found that USL was unlikely to succeed on the merits of its copyright claim because of AT&T's historical distribution practices. He pointed directly to the central fact that Berkeley's defense had raised: the widespread distribution of the original software without a proper copyright notice.

This judicial opinion was a legal "tell" of the highest order. It signaled to both sides that the presiding judge viewed the foundational copyright claim on 32V, the common ancestor of both System V and BSD, as probably invalid. The ruling validated the core of the defense's argument before a full trial had even begun. For USL, the path forward was now fraught with peril. Continuing the litigation meant risking a final, binding judgment from a federal court that would formally declare UNIX to be in the public domain. Such a ruling would have been catastrophic for USL. Judge Debevoise's shadow now loomed over the case, making a full trial a gamble for USL.

A Hasty Settlement to Avert Disaster

The timing of Judge Debevoise's ruling couldn't have been more consequential. In February 1993, just before the damaging opinion was issued, Novell announced its intention to acquire USL from AT&T. The acquisition was completed in July 1993, meaning Novell inherited the increasingly precarious lawsuit. Novell's leadership found itself in an untenable position. They had just spent a significant sum to acquire what they believed to be a valuable proprietary asset. Yet they now faced a federal judge who had strongly signaled that the core copyright on that asset was likely invalid. Compounding this was Berkeley's countersuit in California, which threatened to disrupt Novell's own UNIX System V business.

A definitive court ruling that 32V was in the public domain would have been an unmitigated disaster for Novell. It would have catastrophically devalued the very asset they had just purchased and emboldened competitors. A settlement, on the other hand, offered a way out. It could end the legal bleeding, neutralize the threat of the countersuit, and, most importantly, prevent the question of 32V's copyright from ever being formally resolved. The primary function of the settlement, from Novell's perspective, wasn't to clarify the legal status of UNIX, but to preserve the ambiguity that Judge Debevoise's ruling threatened to destroy. Preserving the copyright claim, even if the underlying legal reality was weak or even nonexistent, was commercially essential.

Novell's CEO, Ray Noorda, faced a judge who openly questioned the existence of the very copyright they had just purchased, and moved to settle.

The settlement agreement, reached in February 1994, is a masterclass in reading between the lines. On its face, it was a compromise. In reality, its terms represent a near-total capitulation by USL/Novell and a validation of Berkeley's position. The key points were:

A New "Clean" Release: The University of California agreed to release a new version of its software, to be called 4.4BSD-Lite, which both parties would agree was "unencumbered" and free of any disputed USL code. The University would encourage users of the previous Net/2 release to switch to this new version.
Minimal Code Changes: The most telling detail is the number of files that were actually affected. Out of the 18,000 files that comprised the Berkeley distribution, USL's claims of widespread infringement boiled down to this: three files would be remove entirely, and a 70 would be modified to display a USL copyright notice. This trivial number stands in stark contrast to USL's initial, sweeping allegations and suggests their case was extraordinarily weak. Even though USL's copyrights on 32V were probably invalid, you can ask for anything in settlement, and Berkeley likely saw that adding notices that were probably invalid would be of little consequence.
Mutual Attribution: In a clear nod to the validity of Berkeley's countersuit, the agreement stipulated that while specific Berkeley files would now carry a USL notice, several USL's files and publications would be required to have a University of California copyright notice and acknowledgment going forward.
Free Distribution Permitted: USL explicitly agreed to permit the free distribution of a list of files defined as "UNIX Derived Files" - code that USL contended was derived from UNIX but which it now agreed could be freely reproduced and redistributed by anyone without a license or fee.

This outcome wasn't the result of a legal victory, but a strategic retreat. A party with a strong infringement claim doesn't settle a lawsuit by demanding the removal of only three files out of 18,000. These terms aren't the result of a negotiation between equals; they're the price Novell paid to make the lawsuit disappear. The settlement allowed Novell to walk away without the public humiliation of a court formally declaring UNIX to be in the public domain. At the same time, Berkeley achieved its primary goal: the ability to distribute BSD without legal threat. The agreement was an act of strategic obfuscation, a mutual decision to bury a legal question that one side was terrified to have answered. The requirement to add copyright notices to a few files was a face-saving gesture, but if the work was already in the public domain, those notices were legally invalid from the moment they were added.

Novell effectively bought a "quitclaim" to the UNIX code - an agreement to end the dispute - while burying the existential threat to its copyright under a (then-sealed) settlement.

The case was over, but the fundamental question it raised remained unanswered. The settlement created a legal ghost: a copyright valuable enough to be bought and sold but too fragile to withstand testing in court. This unresolved ambiguity would lie dormant for nearly a decade before being resurrected in an even larger and more consequential legal war.

The SCO Saga: A Ghost Claiming an Empty Throne

The second half of the story begins with one of the most complex and confusing series of asset transfers in technology history, leading to a legal battle that sought to weaponize the very copyright the Berkeley case had shown to be so fragile.

In 1995, just a year after settling the BSDi lawsuit, Novell decided to exit the UNIX operating system business. It entered into an Asset Purchase Agreement (APA) with the Santa Cruz Operation (also known as SCO), a long-time UNIX vendor. The APA was a masterpiece of ambiguity and the source of all future conflict. The original agreement explicitly excluded all copyrights from the assets being sold to SCO. However, a later amendment carved out an exception for "the copyrights required to exercise its rights" under the agreement. This convoluted language was a legal time bomb.

The chain of custody grew more complex. In 2001, Caldera International acquired the UNIX business from Santa Cruz Operation. A short time later, Caldera changed its name to "The SCO Group". It was this new entity that would launch a legal war against the free software world, based on the claim that it had acquired the original UNIX copyrights through this tangled chain of title.

In 2003, The SCO Group launched its infamous legal campaign, beginning with a multi-billion-dollar lawsuit against IBM. SCO's central claim was that "Linux" was an "unauthorized derivative" of UNIX, containing proprietary code that IBM had improperly contributed. The company then began a campaign to extract license fees from commercial users, creating a cloud of Fear, Uncertainty, and Doubt.

SCO's business model depended entirely on its claim to the UNIX copyrights. When Novell, watching this unfold, issued public statements clarifying that it had retained the copyrights in the 1995 sale, SCO's legal foundation was threatened. In a fateful move, SCO sued Novell for "slander of title," alleging that Novell's public claims were false and were damaging to SCO. This lawsuit, intended to silence Novell, had the opposite effect: it forced a definitive judicial examination of the ambiguous 1995 APA and, with it, the question of who truly held the copyright to UNIX.

The SCO v. Novell litigation dragged on for years, through summary judgments, appeals, and finally, a jury trial. Finally, the verdict was delivered: a federal jury found unanimously that Novell, not The SCO Group, held the UNIX and UnixWare copyrights.

The court's interpretation of the APA was decisive. It sided with Novell's argument that the convoluted language of the agreement had not, in fact, transferred the copyright. Novell had retained it to protect its ongoing 95% interest in licensing royalties and had merely granted SCO's predecessor a license to use the code.

Crucially, the court's decision was retroactive. It didn't revert ownership of the code from SCO back to Novell; it declared that the copyright had never been transferred in the first place. This meant that, for all the years that Caldera and The SCO Group had been acting as the copyright holder - filing lawsuits, selling licenses, and threatening the entire free software community - they had been asserting rights they never had. Every action they took as the purported copyright holder was legally void from the beginning. SCO's own legal aggression had forced a clarification that proved its entire business model was built on something imaginary.

We're now left with two seemingly contradictory legal histories. The USL v. BSDi case cast serious doubt on the validity of a U.S. copyright for early UNIX. The SCO v. Novell case, years later, definitively ruled on that same copyright. Juxtaposing these two outcomes reveals a profound legal paradox, one that leaves a trail of unanswered questions.

Here lies the central paradox: In 1993, a federal judge strongly suggested that the copyright for UNIX/32V was likely invalid in the United States because it was published without proper notice. In 2010, another federal court declared Novell the undisputed holder of the UNIX copyrights. How can a company be declared the holder of something that may not exist?

The answer lies in the different questions each court was asked to answer. The USL v. BSDi case involved a motion for a preliminary injunction, in which the judge had to assess the validity of the copyright to determine whether an injunction was warranted. The case was settled before a final, binding ruling on that specific question was ever made. The SCO v. Novell case, conversely, was a "slander of title" dispute. The court's task wasn't to determine whether the copyright was valid, but to decide which of the two parties - SCO or Novell - held the title to whatever rights existed. The case proceeded on the assumption that there was a copyright to be held.

This distinction, while legally precise, does little to resolve the practical paradox. Novell won the the deed to a house that a previous inspector declared may have been built on a sinkhole. Novell is the confirmed holder of a U.S. copyright that, for early versions like 32V, may not exist. This leads to the next logical question: what of the licenses granted by the entity that never had the copyright?

In 2002, Caldera International - the entity that would become The SCO Group - released several "ancient" UNIX versions, including V1-V7 and 32V, under a permissive, BSD-style license. This act firmly placed this historic code within the free software ecosystem.

However, does the retroactive ruling in SCO v. Novell completely undermine this action? One can't license rights one doesn't have. A license is a promise from a copyright holder not to sue for infringement. If the entity granting the license isn't the copyright holder, that promise is legally meaningless. Since the ruling established that Caldera/SCO never had the UNIX copyrights, is the license it granted in 2002 void ab initio - was it invalid from the moment it was issued?

This raises a question: If the permission granted by that license is a legal fiction, on what basis can anyone use, modify, or distribute this historic code?

When all the pieces are assembled, an argument emerges that the ancient versions of UNIX are probably, for all practical purposes, in the public domain within the United States but you'd need an attorney to make that case. This status arises not from a clean, explicit dedication, but from a confluence of legal blunders and historical accidents.

No court has ever issued a final judgment explicitly stating that ancient UNIX is in the public domain. However, to focus on this absence is to miss the forest for the trees. The chain of events surrounding the USL v. BSDi lawsuit, when analyzed as a whole, constructs a persuasive circumstantial case for precisely that conclusion. The settlement wasn't a resolution of the question, but a pragmatic business decision to ensure the question was never formally answered.

Finally, the actual copyright holder, Novell, has gone on record stating that it has "no interest in suing people over Unix" and that "We don't believe there is Unix in Linux," signaling an apparent lack of intent to enforce any rights it may hold.

Does this combination - a likely forfeited copyright, a voided license, and an uninterested copyright holder - create a "de facto" public domain status?

And if so, we are left with the ultimate question: Is this de facto status legally sufficient? Without a definitive court ruling declaring the work to be in the public domain, or an explicit dedication from the confirmed copyright holder, a sliver of legal uncertainty will always remain.

In the end, we return to where we began, with the ghosts in the machine. The legal battles are over, but they have left us with a paradox. One court case definitively settled who had the throne. But an earlier, unsettled case suggested that the throne itself might be an illusion. Novell won the deed, but the property may have been condemned long ago. A pretender wrote the license that promised freedom.

The story of UNIX's copyright is a true ghost story: a tale of a copyright that may or may not exist, yet has driven acquisitions, litigation, and corporate strategy. We may never get a definitive answer, leaving the ghost of ancient UNIX copyright to wander the halls of computing history, a permanent mystery.