Inside the Linux arcana

commentary Did a 1994 agreement with SCO hand Sun a smoking gun?
Written by David Berlind, Inactive
commentary Did a 1994 agreement with SCO hand Sun a smoking gun?

If the SCO Group has any hope of prevailing in its core cases with Novell and IBM, those hopes are pinned on the following findings sequentially falling into place: establishing what Unix intellectual property (IP) it owns, to what extent it owns the IP (the Novell case may prove that it's shared in some way), what enforceable rights the company is entitled to by virtue of that ownership, and then how those rights may have been violated.

Considering all the chips that must fall in SCO's favour, and the order in which they must fall, the company has several milestones to get past before it can begin singing, "we will, we will rock you." It's a slippery slope, and the legal experts I've interviewed agree that the tipping point may reside in a sealed settlement to a 1992 case that was brought against both the University of California at Berkeley and Berkeley Systems Design (BSDi) by Unix Systems Laboratories (USL), an AT&T spinoff of which AT&T was still the majority stakeholder.

Although the settlement is sealed, one of the most widely held beliefs about it is that it privately acknowledges a significant dilution of AT&T's copyrights and trademarks because of the contributions to Unix that were made by UC Berkeley. Should the settlement be unsealed and the language proven to undermine the integrity of the SCO's IP claims, SCO's case could be significantly weakened - if not finished.

In addition, to the same extent that the sealed settlement may call into question the integrity of the IP in question, it could also turn out that a 1994 agreement between Sun and USL (very shortly after USL was acquired by Novell) holds more evidence of the same sort of IP dilution. It's just a hypothesis at this point, but it could turn out to be a powerful "smoking gun" that influences the outcome of the various suits brought by SCO. Before further addressing the issues of sealed documents and potential secret deals, it's useful to outline current events and to go over some background history.

The cases have a lot of milestones and, as indicated by the orders issued two weeks ago by US District Court (Utah) Magistrate Judge Brooke Wells, those milestones may not appear to follow the logical order that I have just described. As part of the due diligence needed for the case against IBM to proceed, SCO must show the court a smoking gun -- Unix-related code from which Linux was improperly developed or derived and memos from IBM executives that indicate complicity with, if not authority over, certain Linux development decisions. If the judge were to see enough smoke emanating from the gun's barrel to proceed with the case, the pendulum will have swung in SCO's direction, but it does little more than guarantee that the company will get its day in court.

In her 2 March, 2004, order, Judge Wells directed SCO "to identify with specificity all lines of code in Linux that it claims rights to." She hasn't yet asked SCO to substantiate those rights. If the code is eventually proven not to be SCO's, or if it is but the court doesn't side with SCO on its enforceable rights against Unix licensees (both significant milestones), then the code itself may end up being irrelevant to the lawsuits against IBM.

The relevancy of other alleged code-related improprieties and the degree to which SCO has enforceable rights over non-Unix licensees such as Linux-kernel chaperones Linus Torvalds or Andrew Morton, may be different. So far, SCO hasn't sued them.

BSD: The potential short circuit to SCO's case
If there's an opportunity for the cases to proceed out of order, it's because the IP rights (IPR) that go with Unix might never have been enforceable in the first place. In other words, the copyrights might not have been enforceable, even back in the 70s and 80s when AT&T owned them. If AT&T's IPR wasn't enforceable, then it stands to reason that the IPR of all other subsequent owners (Unix Systems Laboratories, Novell, the old SCO, Caldera, and now the new SCO) weren't enforceable either.

If the IPR turns out to be useless, it will turn out that Novell bought the Brooklyn Bridge as well and subsequently passed it along to SCO. Worse for SCO, if there's essentially no IPR to own, then the ownership issues underlying the suits are moot.

As I noted earlier, the answer to the question of enforceable IPR may reside in a sealed settlement to a 1992 case that was brought against both the University of California at Berkeley and Berkeley Systems Design (BSDi) by Unix Systems Laboratories (USL).

By 1992, almost two decades had passed since UC Berkeley, AT&T's Bell Labs, and others began an open source-like collaboration on the development of Unix. According to a passage written by Marshall Kirk McKusick (from the book Open Sources: Voices from the Open Source Revolution), who was deeply involved in Unix development during its embryonic years, "BSDi was formed to develop and distribute a commercially supported version of [BSD Unix]." The initial version was almost entirely based on the most recent release (at the time) of a version of BSD Unix known as Networking 2 (Net/2) that had advanced networking features (for the time) and that its developers believed to have no USL-based encumbrances.

Up until that point, the only commercially available versions of Unix came from USL licensees. According to McKusick, the USL lawsuit was prompted by BSDi when it "began selling their system including both source and binaries in January 1992 for US$995. They began running advertisements touting their 99 percent discount over the price charged for [AT&T's Unix] System V source plus binary systems. Interested readers were told to call 1-800-ITS-Unix."

Had the case been settled in court, it might have set the precedent for what constitutes an illegitimate derivative work, a precedent that might have been relevant to the current cases involving SCO. Net/2's roots can be traced back to a set of TCP/IP network utilities that were developed without USL's involvement, but were used to network-enable BSD, an operating system that still contained AT&T intellectual property. According to McKusick's book, UC Berkeley started with those core networking utilities and worked backwards through the rest of BSD, slowly replacing any old code to which the USL copyrights applied.

SCO maintains that the practice of using code (for example, the networking utilities) that wouldn't have existed in the first place had it not been for the existence of an original core (the AT&T-encumbered version of BSD), and then working backwards from that point to replace the original core so as to arrive at a completely new and unencumbered product, is a disallowed form of derivation. Such a process is not the basis of an SCO claim against UC Berkeley, but rather, IBM. From a market point of view, the existence of BSDi (and its marketing language) proved how such a derivation process could result in an offering that undermines the business of the IP holder to the original core.

USL must have seen it that way as well, suing on the grounds that the release of the BSD operating system resulted in the exposure of its trade secrets while also violating its copyrights and trademarks. BSDi was targeted as the commercial offender while UC Berkeley earned the dubious honour of being sued because it was the primary source of the code used by BSDi to create its commercial releases. USL was looking to reign in what it saw as a significant threat to its business. However, there was one hitch.

In an Open Source Initiative position paper regarding SCO's complaint against IBM, OSI president Eric Raymond argues that a judge's failure to order an injunction that would have favoured the plaintiffs might have been an early indicator that due to substantial contributions made to Unix by UC Berkeley, the court was unwilling to acknowledge the enforceable rights to Unix that USL was claiming it had.

The case was settled in January 1994 and, though the terms are sealed, the bartering between the plaintiffs and the defendants yielded an awkward situation in which the AT&T/USL copyrights stayed with a relative handful of the approximately 18,000 files in the settlement-driven successor to Net/2: 4.4BSD-Lite. Though 4.4BSD-Lite contained USL's copyrights, UC Berkeley (technically, the Regents of the University of California) was given expansive rights to use and distribute it in BSD without interference from the USL or its successors (such as SCO). One of SCO's complaints is that some of that BSD code found its way into Linux.

Amongst all distributions of Unix, BSD has been the only one to get such access. The arrangement is best summed up by a passage found on the OpenBSD Web site that says, "while AT&T holds the copyrights to much 'Unix' code and documentation, OpenBSD is based largely on Berkeley (BSD) distributions that contain only material known to be free of AT&T copyrights, or material to which AT&T has abandoned its copyright or included licensing terms similar to the Berkeley terms. No material subject to restrictive AT&T copyrights can be included in OpenBSD." OpenBSD is a BSD variant with an emphasis on security.

In response to any assertion that the sealed settlement reveals the unenforceability of the copyrights to Unix, SCO officials contend that the continued presence of the USL copyrights on the BSD source code is proof that the IPR is enforceable, rather than unenforceable as suggested by Raymond. SCO maintains that UC Berkeley acknowledged their enforceability by agreeing to keep the copyright with the code. This way, no other disallowed derivations -- such as Linux, from SCO's viewpoint -- could come out of the woodwork.

SCO's argument seems logical. However, the same could be said of the counterargument, which maintains the settlement was a barter agreement whereby both parties got what they wanted. Berkeley got a hall pass to do what it wanted with BSD, while the continued presence of copyrights on its source code allowed Novell (which had acquired USL by this point) to avoid an uprising by Unix licensees who, unlike Berkeley, were paying dues and might decide not to pay should the IPR's unenforceability come into question. And, so goes the counterargument, to protect Novell's Unix business the settlement was sealed. None of the involved participants are allowed to discuss its contents.

So far, the potential of the sealed settlement to confirm or deny the enforceability of the original IPR, perhaps short-circuiting the entire case, is nothing new amongst the cases' most fervent observers. What hasn't been discussed is whether another "sealed" document that could be equally damaging to SCO's case might exist. For example, a document that, on the basis of other contributions made to AT&T's Unix System V, gives the contributor(s) rights to1 the Unix IPR that are equivalent to or perhaps even more expansive than the access that UC Berkeley acquired in its settlement with Novell. If such rights were awarded to yet another contributor besides UC Berkeley, then those rights could end up shattering SCO's claims.

I'm almost certain that such a contributor exists -- Sun Microsystems. Although the company won't officially confirm or deny my theory, I came across some old notes in which Jonathan Schwartz, Sun's executive vice president for software, discussed the "extremely expansive" rights to Unix that the company acquired 10 years earlier, in 1994.

Sun has declined to comment on just how expansive those rights are, but the negotiations for that license were impeccably timed with the most gut-wrenching phases of the litigation over Unix, as well as the acquisition of USL by Novell.

Perhaps starting with the contributions to Unix made by Sun co-founder Bill Joy before Sun even existed (going back to 1975), the histories of Unix and Sun are deeply intertwined. Though many people at Berkeley contributed to the evolution of Unix, McKusick's treatise on the history makes it clear that Joy was one of the dominant figures in Unix's early days. Mention of his pioneering work on the operating system's kernel, interprocess communications, and text editors barely scratches the surface of his overall contribution. Joy may have left Berkeley to start Sun, but he was hardly finished influencing Unix (or, at the very least, a variant of it).

Moving forward from the early 1990s, a variant of Unix known as System V Release 4 (SVR4) would become the basis of Unix licensing by AT&T and its successors. As the closest thing to a successful standard for Unix, most derivations of the operating system ended up being SVR4-compatible. But, virtually every version of the SVR4 family tree reveals an interesting root: Sun's version of Unix known as SunOS, which not surprisingly (because of Joy's work at Berkeley) could trace its roots back to BSD. Most characterisations of SVR4's evolution cite the collaboration between AT&T and Sun (under the heading of Unix International) that merged technologies from BSD, AT&T's SVR3, and Sun's SunOS.

In 2002 I interviewed Rob Gingell, Sun's chief engineer, who characterised the contribution as substantial when he said: "Just about everything that started the Unix wars, which was based on AT&T picking Sun OS as the basis for going forward with Unix, everybody now has it."

Although I've been unable to get confirmation on the code names, several roadmaps that can be found on the Internet show how SunOS family tree sprouted the branch between the time it released SunOS 4.1.1 (known as Solaris 1 and also code-named Zeus) in November 1990 and when it released the 4.2BSD-based SunOS 5.0 (known as Solaris 2.0, code-named Jupiter) in July of 1992.

The timelines suggest that Sun's work on Jupiter, and probably Mars (which Sun intended to be a fully SVR4-compliant version follow-up to Jupiter), was flowing directly into SVR4. Among the technologies swimming downstream were Sun's Network File System (NFS), OpenLook (a graphical user interface), X11/NeWS and some memory management technologies.

According to Sun folklore, computer tapes (with the code) were sent via FedEx to AT&T in New Jersey almost every Thursday night. Meanwhile, as AT&T was putting the finishing touches on SVR4, Sun moved forward with its plans to release SunOS 5.0 (Jupiter) and then, five months later in December of 1992, released the fully SVR4-compliant SunOS 5.1 (Solaris 2.1, aka Mars). For history buffs, while all this was taking place, Sun's Bill Joy and James Gosling were already working on Oak, which later became Java.

By now you might be asking, "what are the implications relative to SCO's legal claims?" Schwartz's usage of the term "expansive" to describe the rights to Unix that Sun acquired in 1994 is a fairly strong hint that its rights go beyond that of a typical licensing agreement. If I were on Sun's executive team in 1993, and I saw how, on the basis of the contributions to Unix made by UC Berkeley, the code was about to become permanently unencumbered by USL's IPR, I would be asking the following question: if UC Berkeley's contribution to Unix essentially stripped USL of its enforceable rights, couldn't the same be said of Sun's contribution to SVR4? In light of that, knowing that Novell's possession of USL (and its IPR) offered no assurances as to where the IPR might one day end up, now would be a good time for Sun to permanently secure some rights to Unix that, as a contributor, were equally if not more expansive than the ones UC Berkeley was getting. After all, who knows what hands the IPR might eventually end up in and what sort of aggressively litigious posture they might take?

Did Sun have an uncanny bout of clairvoyance and act on its intuition by negotiating for itself the same sort of sweet deal that Berkeley got from the settlement with Novell? Beyond Schwartz's characterisation (offered in a context outside of SCO's lawsuits) of the 10-year-old transaction as a "buyout of Unix" that involved "expansive rights," Sun officials haven't elaborated. Based on historical accounts of Unix found on the Web and even a recent News.com report, Novell's buyout cost to Sun was US$82 million and took place in 1994. Beyond reports of a perpetual license, not much is known about the finer details of the 1994 transaction.

Although this interpretation is purely speculative, one reason for keeping the terms quiet could be the same reason that some believe the Berkeley settlement with Novell is sealed -- any public admission by Novell that Unix was a community effort involving contributions by not just Berkeley, but Sun and perhaps others could render its just-acquired (for US$800 million) Unix IPR useless. Paying licensees would put their checkbooks away.

Already in the midst of the settlement with Berkeley, if Novell feared such dilution and Sun knew about it (and it must have had some great inside connections at Berkeley), then Sun would have had the upper hand in any negotiations. It also had the motive to use it. Reports, or perhaps urban legends, that I found on the Web paint a picture of a Sun that was very dissatisfied with the prospect of Novell owning Unix. Keeping the agreement stowed away under Sun chief executive Scott McNealy's pillow would hardly be considered a sacrifice if, in return, Sun ended up with an unrestricted license in perpetuity.

Ten years later, while many eyes are still trained on the sealed settlement with Berkeley as a potential short circuit to the enforceability of SCO's Unix IPR, no one appears interested in the deal that Sun may have cut for itself, which could have equally if not more damaging implications for SCO's case than the Berkeley settlement.

One reason might be because of the news report about how, in February 2003, Sun cooked a deal with SCO that "expanded" the rights it acquired in 1994. According to News.com reporter Stephen Shankland, Sun spokesman Brett Smith confirmed a transaction with SCO whereby Sun is now permitted to "use some software from Unix System V Release 4 for software components called drivers, which let computers use hard drives, network cards and other devices. Sun needed the software for its version of Solaris that runs on Intel servers."

SCO's version of the story, which appeared in one of its SEC filings, said: "We initiated the SCOsource effort to review the status of these existing licensing and sublicensing agreements and to identify others in the industry that may be currently using our intellectual property without obtaining the necessary licenses. This effort resulted in the execution of two licence agreements during the 30 April, 2003, quarter. The first of these licences was with Sun Microsystems, Inc. ('Sun'), a long-time licensee of the UNIX source code and a major participant in the UNIX industry, and was a 'clean-up' licence to cover items that were outside the scope of Sun's initial UNIX licence."

Shankland's report also says that the transaction, according to a regulatory document, "granted Sun a warrant to buy as many as 210,000 shares of SCO stock at US$1.83 per share as part of the licensing deal." SCO's most recent disclosure claims there are 14.4 million shares outstanding. Any exercise of those warrants would constitute a better-than-one-percent stake in the company, a stake that could be significant to the fate of the IPR should SCO have to close its doors.

Indeed, some money may have exchanged hands back in February 2003. But, I wonder if publicly offered explanations for the licence clean-up and transaction were less than full disclosure.

My theory (pure speculation) goes like this: if the rights that Sun acquired for US$82 million in 1994 were so expansive, and given the engineering resources that Sun has at its disposal, the mere suggestion that Sun needed to license some driver technology from SCO for the x86 versions of Solaris seems ludicrous. Even if Sun's developers were too busy to develop the drivers on their own, all they had to do is pick up the phone and call Intel or AMD and a CD-ROM would have been in the mail the next day. In an effort to expand their customer base, microprocessor companies like Intel and AMD are usually happy to provide whatever development resources are necessary to seize a new opportunity.

In an interview, Intel chief executive Craig Barrett said that he would welcome the opportunity to work with Sun. Said Barrett of McNealy, "If he wants -- and he's serious about a design win with us -- we are willing to talk to him."

Of course, later that year at Fall Comdex Sun entered into a strategic partnership with AMD.

If the need to expand its already expansive licence was dubious at best, what could Sun's motivations for visiting SCO in Utah have been? For starters, consorting with SCO makes Sun's 1994 buyout look less interesting to SCO's opponents. Another reason for Sun's visit, especially so early in the timeline of SCO's legal pursuits, could have been to show SCO that it had that card in its back pocket. If you were a fly on the wall, you might have heard something like, "if you think the Berkeley settlement was bad, wait until you see this. If you even sneeze in our direction, you'll leave us no choice but to show our hand. So play nice and give us some warrants in case you go bankrupt. That way, if there's any residual value to the Unix IPR, we have a say in what happens to it."

Of course, a bigger question underlying my hypothesis is why hasn't Sun publicised more of the details from the 1994 agreement. If Sun really has a smoking gun, it would clearly have an adverse impact on SCO's legal manoeuvres. It could be said that, like Microsoft, Sun has an interest in seeing SCO win. After all, Windows isn't the only operating system that has suffered at the hand of Linux. It has taken a bite out of just about every Unix-licensee's business (including Sun's).

If SCO prevails, and Linux is no longer distributable under the GPL, one of Linux's most competitive qualities versus other operating systems, which require a paid licence, goes away. Instead of blowing up SCO, perhaps Sun is doing the opposite -- helping underwrite SCO's legal pursuits but couching the transaction in terms that most people would find to be plausible. This might sound familiar to anybody that's been following Microsoft's money as well.

To be clear, we might never know whether Sun went to Utah to scare SCO, to indirectly underwrite its legal endeavours, or simply to "clean up" things. But, whereas Linux's continued availability under the GPL or lack thereof (a fate which most certainly will be determined by the outcome of the SCO case) is relevant to the future of Linux distributors like Red Hat, Sun's Schwartz implied in an email exchange with me that the case and its outcome have no bearing on Sun's decision-making or it plans to win with Solaris.

The reason Schwartz appears to have no interest in the case is because he claims that Solaris x86 isn't competing against an army of GPL-emboldened Linuxes. Instead, given Red Hat's overwhelming dominance on the enterprise Linux front, Schwartz says Solaris is competing against Red Hat Enterprise Linux (RHEL). Indeed, according to IDC, Red Hat is the dominant supplier of Linux servers, coming in with more than a 60 percent share of the market. According to Schwartz, between RHEL's lack of binary compatibility with other Linux distributions, the overwhelming demand for it, and the fact that Red Hat now charges for it, the GPL has been neutralised from both the enterprise buyer's point of view as well as from Sun's competitive point of view. As far as he's concerned, the business playing field has been levelled and he's competing with an operating system that might as well be proprietary. Red Hat apparently sees the same thing in reverse, viewing the Unix operating system as the company's main competition.

Against that sort of competitor, and armed with Sun's engineering resources as well as a new licensing scheme, Schwartz is confident that Sun can win, GPL or no-GPL. "We're continuing to invest in our OS, Solaris -- running on Opteron, Intel and Sparc," said Schwartz. "So we've got a lot more flexibility to continue driving improvements, and price/performance. That's why you'll see Solaris on x86 systems priced lower than Red Hat." To address Red Hat's subscription model, Schwartz said, "soon, we'll announce a subscription-based pricing model that will allow you to compare apples to apples with Red Hat."

The outcome of the SCO case and whether the GPL is bound or unbound from Linux may indeed be immaterial to the Sun's game plan. But, I'm not certain that Sun wouldn't be just as happy to nudge Red Hat off the cliff if it got close to the edge. I'm reminded of the fact that Schwartz was the one who pointed out Red Hat's 10-Qs to me when I asked whether Red Hat was the canary in SCO's coalmine.

In the sections that address SCO, those 10-Qs make it clear that Red Hat might be unable to continue selling its products if they could no longer be distributed under the GPL. Schwartz may indeed be confident that it can beat Red Hat and its hardware partners if the GPL remains bound to Linux. But, I find it hard to believe that he and other Sun executives wouldn't be downright giddy if SCO prevailed and Red Hat subsequently imploded.

Of course, this is all speculation, but the trail of facts seems to indicate that Sun has some information or an agreement that could impact the outcome of SCO's legal manoeuvres. At the same time, many questions remain unanswered. For example, if Sun had so much leverage, why did it have to pay US$82 million to Novell in 1994 for its expansive rights? With a US$800 million deal hanging in the balance, did Sun strong-arm AT&T -- its Unix partner in kind for over a decade -- before USL was sold to Novell? Perhaps the answers to these questions lie under Scott McNealy's pillow, and perhaps not. We'll have to wait to see how it all plays out.

Editorial standards