SCO, Groklaw and the Monterey mystery that never was

Bullets for SCO?

Unix before Linux

In October 1998, when IBM plumped for a joint development agreement with the Santa Cruz Operation and Sequent, it was a bold move. The past thirteen years at Big Blue had been bedeviled by two JDAs which had proved catastrophic for the company.

In 1985 IBM signed a JDA with Microsoft to develop OS/2. It ended the agreement in acrimony with Microsoft holding the trump cards, in 1991. That same year, IBM teamed up with Apple to produce a next-generation operating system, on whose fortunes IBM's proprietary operating systems would eventually converge. Taligent and WorkPlace OS "are evolving in parallel, and IBM and Taligent are trading technologies back and forth along the way," reported Byte in 1994. The expensive venture yielded nothing, and a year later, Apple stubbornly refused to be acquired by IBM at the eleventh hour.

With these memories fresh, IBM embarked on a plan to merge its AIX Unix for RISC with SCO's UnixWare and Sequent's Dynix/ptx Unixes for x86, porting the result to Intel's forthcoming 64-bit processor, code-named Merced, and licensing the result to all-comers. Were they crazy? Not quite.

In 1997, when the plan was being hatched, analysts and the big iron vendors braced themselves for Merced to be a roaring success, making proprietary RISC processors like IBM's own PowerPC uneconomical to produce. Several RISC vendors struck OEM deals with rivals - for example, SGI spin-off MIPS sold its processor to Siemens and Tandem as well as SGI, and Hewlett Packard sold PA-RISC to Hitachi, Stratus and NEC - but none could generate volumes as the Intel world understood volumes. SGI and Hewlett Packard had declared that the unreleased Intel chip was the long-term future. (The instruction set was unveiled in October 1997 and branded Itanium® in 1999).

So the fear factor played a major part in IBM's decision, but opportunity beckoned too. IBM had shunned industry standards in the 1980s when it introduced its PS/2 series with a proprietary bus, and was aware that licensing allowed it to reach a potentially greater market. Meanwhile, IBM prudently continued to invest in its Power RISC microprocessor architecture.

For the Santa Cruz Operation, a different kind of opportunity beckoned: the big time.

SCO had been licensing its venerable OpenServer Unix for Intel, formerly Xenix, with steady success for a decade, and OEMs ranged from tiny VARs producing point of sales systems to the biggest PC vendor, Compaq. OpenServer was ubiquitous in retail and distribution, and SCO's unselfish approach to business had encouraged a loyal channel. But OpenServer didn't scale, so in 1995 the company acquired something that could, UnixWare from Novell.

However, for SCO to develop UnixWare into a serious enterprise contender required a partner with deep pockets. HP flirted with SCO briefly, and when that fell by the wayside, SCO talked to Sequent and Compaq about standardizing interfaces with their joint Bravo Unix project. This was based on DEC's OSF/1 but ported to run on Intel's Merced and Compaq's Alpha processor. The hope was that UnixWare and Bravo would be binary compatible.

But when IBM itself came knocking, it looked like the perfect fit for SCO. With the potential of the 64-bit Intel market, and IBM's mighty Unix division as a partner, SCO looked like it could hit the big league, with an enterprise class Unix selling on a high-volume processor.

Sequent, who like SCO had been selling Unix on Intel since the early 1980s (only at the high end with multi-processor machines) switched camps, and so in October 1998 Project Monterey was born.

Meanwhile SCO was aware that a low cost Unix on Intel posed a potential threat to its OpenServer business, in the shape of Linux. The OS had been enthusiastically adopted in universities and provided a cheap and dependable platform for network infrastructure. It was no contender to OpenServer, but that changed when a company called Caldera emerged in 1996. Caldera respectfully mimicked SCO's channel-based business model, and declared SCO customers to be low hanging fruit. By 1998 Linux was gaining publicity, and SCO executives were increasingly fielding questions about something that didn't look like a competitor as much as a movement.

SCO made friendly with Linux as best it could, pointing out that at the high end, UnixWare was scaling up to 30 node clusters in the labs, and promising it would keep OpenServer up to date as long as customers wanted it. It also demonstrated a way of running Linux binaries on UnixWare, the Linux Kernel Personality (LKP) project.