The Love of Linux Chapter 12: The Formation of Standards-UnitedLinux; a New Paradigm of Collaboration
One of the greatest principal to be learned from Linux and the formation of UnitedLinux is the power of a new type of collaboration even among competitors. Caldera was very instrumental in both the formation and the down fall of UnitedLinux but the principals behind the formation of UnitedLinux live on and can offer great advantages to companies who embrace them. To understand Caldera’s motivation for the encouraging UnitedLinux, one needs to understand the need to establish a platform supported by Independent Software Vendors (ISVs) and Independent Hardware Vendors (IHV).
Volume attracts developers and standards help to increase volume and to gain broad market acceptance of the ISVs and IHVs. If these developers know that the same interface or method of doing something is used not only by one company but all companies utilizing that technology, they can reach a far bigger market by developing to that interface or standard. Caldera, almost from its inception, has been very vocal and financially committed to the support for standards, recognizing that if Linux were fragmented ISVs and IHVs would have a very difficult time supporting Linux over time. It is important to understand a brief history of standards within the Linux movement. Caldera was committed and played a key role throughout this process because we understood the importance of the support of these ISVs and IHVs.
Shortly after the formation of Linux International, an influential developer in the Linux Community named Bruce Parens approached the Linux International board with the idea of beginning a standardization effort for Linux that included a reference port for Independent Software Vendors. Caldera was very vocal about the need to have a standard and a reference implementation. Bob Young of Redhat was supportive of a standard but did not want a reference port. After some rather contentious email discussions, Bruce Parens stepped out of the effort and the open source movement for a time.
Linus Torvalds intervened and almost self appointed a coworker Dan Quinlan to head up the effort. About this time, Redhat partnered with Debian to establish an independent standard, but after an outcry from the Linux community, RedHat and Debian conceded and agreed to support what had come to be known as the Linux Standard Base effort. Dan spent a great deal of time lobbying the different groups in garnering support for the standard. Caldera’s lead Linux engineering manager, Ralf Flax was nominated to head up the reference port implementation. Progress was slow and painful. Those who were often the most vocal on the email groups were interpreted as being supportive of the effort when in fact they appeared to be stalling the process.
The feeding frenzy that took place as several Linux companies went public greatly slowed the attention paid the standards efforts. The spotlight of the industry was on Linux and it seemed as though the Linux companies could do no wrong. Every leading OEM and ISV was jumping on the Linux bandwagon and wanted to be associated with Linux. Investments were being announced; marketing and support agreements were being signed. Geographic regions of market share were being carved out and everyone was racing to pull ahead of the others. Significant amounts of capital were being poured into the industry and duplicate infrastructures were being created. Each was trying to leverage their geographic leadership to gain global dominance. In this over heated environment, some in the community where not as motivated to come together and cooperate, because they all felt that they could be the number one. In addition, in fairness to all, resources were limited and the rapid release cycle the open source business model required, greatly constrained the amount of time and resource the Linux companies could allocate to a cooperative effort when they were racing to gain market share and presence. Many were claiming the ultimate victory of open source as the new way for software to be developed, marketed and sold. Linus and Dan’s company went public. Dan stepped down as the head of the LSB effort. The former President of SuSE in North America, Scott McNeal picked up the pieces and formed the Free Standards Group and brought the Internationalization standards and LSB under one umbrella. The specification LSB 1.0 and 1.1 was finally completed and released.
Then the financial market came crashing down and with it the reality that unless there was a business model to sustain the ongoing development and global infrastructure, Linux companies could not survive, let alone expand. As the investment dollars dried up, all the Linux companies began to shrink and return to their own areas geographic areas of strength. As the economy slows, the major players cannot continue to support multiple versions of Linux. As discussed previously, the costs of certification are not insignificant. The retail market followed the economic downturn as a result of the .com implosion and Caldera’s revenues were directly tied to retail sales. Caldera inherited two additional operating systems from SCO, OpenServer which was driving the majority of the revenue and UnixWare, a System 5 Release 4 version of UNIX on Intel. These operating systems both had significant, but unique challenges. While the OpenServer operating system had wonderful legacy application and hardware support in the retail market, it was very old and lacked the architecture and support of the new hardware platforms. New application developers were looking toward the Internet and windows to seek the elusive dollar. OpenServer had little or no Internet functionality and needed a significant overhaul to be able to compete in the new markets. Linux was an ideal substitute for OpenServer because it required very little hardware resources, it had all the needed Internet functionality; it was extremely stable and very low cost.
All Linux lacked was the legacy driver and application support of OpenServer to significantly penetrate the retail market. SCO’s management could have been the dominate leader in Linux if they had embraced it early on. However, they could not see a way of making money with Linux and feared that it would erode their revenue stream. Instead, they bought UnixWare from Novell. Unfortunately, UnixWare was very much overkill for the majority of SCO’s market. UnixWare required significantly larger servers with larger disks and memory, making the overall solution for the small business environments too expensive to deploy. Besides, Novell had already mined out the Value Added Resellers (VAR) that were willing to carry UnixWare. The VARs and distributed enterprise customers loyal to OpenServer could not afford to deploy UnixWare. I can only speculate why SCO management choose to purchase UnixWare rather than investing in Linux. 1) SCO management could not figure out how to make money on Linux. I had several conversations with Doug and Alok on this topic so I know that it eluded them. In their defense, the successful models described above were only visionary at the time and there were no one who had yet proved them 2) SCO Management wanted to play with the “big” boys because it appealed to their egos
Everyone wanted to be the “enterprise” player to be the next “Microsoft” rather than develop a strategy of their own. I watched Novell almost grow to disdain the very small business channel that made and kept them in business for so long. Novell loved the revenue generated by the cash cow “Netware” but almost resented the fact that it was “small” business that made the money. I believe that SCO suffered from this same desire to “arrive” in the big leagues. 3) SCO management believed the Intel marketing hype and thought the new generation of Intel platforms would actually allow them to play in the enterprise. In many ways, UnixWare was even overkill for the Intel architecture. Many of the enterprise features UnixWare contained could not be deployed on Intel hardware platforms and when they could, few if any major corporation seriously considered deploying Intel in the true enterprise. Intel Architecture (IA) 64 bit was going to change all of this. While “enterprise” feature rich, UnixWare never had enough volume to attract either hardware or software developers. It was the epitome of a wonderful technology waiting for a market. Unfortunately, although UnixWare was in fact mainstream UNIX functionality, it significantly lacked Internet performance and functionality. SCO was relatively a small company. They did not have the resources to maintain, let alone enhance, two Operating system platforms. Furthermore, they did not have the resources to try to “buy” the support of the OEMs, ISVs and IHVs for two platforms since they both lacked current market support.
The SCO management had literally painted themselves into a corner because the ignored the market and their customer. The SCO management had failed to upgrade OpenServer, ignored Linux and the Internet (textbook Innovator’s Dilemma) and purchased a product that was of relatively little value to their current customer base. From the long term (20 year) lease agreements we had to buy our way out of, I firmly believe they also completely misinterpreted the increase sells that occurred around the turn of the century. Many companies purchased new licenses to avoid the dreaded “Y2K” fears. The SCO Management probably interpreted the increase volume as market acceptance of their very poor choices and began to act like they were on top of the world only to watch the bottom fall out of their market and their stock price within the first few months of 2000.
While we had an idea of what we were getting into when we purchased SCO, we did not anticipate fully the predicament we now found ourselves in and the deteriorating economy amplified significantly. We had limited cash, an obsolete operating system (OpenServer) that was bringing in the cash, expenses that far exceeded revenues, and revenues that were shrinking monthly with the economy. We also had a “primer” Operating system (UnixWare) that needed applications, hardware and Internet support, and customers and a Linux operating system that lacked endorsement by the major ISVs and IHVs because of the increasing costs to certify. RedHat was quickly becoming the default certification platform for Linux. If the rest of the Linux companies could not provide a unified platform, the ISVs and IHVs would have just gone with RedHat.
As the market started to deteriorate, Caldera and several others began to have discussions on collaborating on an LSB reference platform. John Terperstra, a prominent contributor to the SAMBA development effort and well respected in the open source community, left Turbo Linux and joined Caldera. John began to have successful discussions with representatives of SuSE, Turbo and Conectiva. Now adverse economic conditions hampered the success. Development resource cycles were even scarcer. At Caldera’s Forum in August in 2000, I made a public plea that it was time to come together and form a Common Business Linux effort. Intel and HP began promoting Debian as a way to standardize around the LSB specification.
From August, Caldera began working to get traction with the major Linux providers. Turbo was highly supportive of the effort, but SuSE was distracted by interval restructuring efforts. Rick Becker, from Compaq arranged a meeting in Munich to begin the discussion in earnest. SuSE had an interactive management team, and discussions fell apart. Caldera then went to IBM and a meeting was set up in Barcelona, Spain. SuSE sent a team there under the direction of the interim management team. We nearly had to break of negotiations because we could not get past some very key points. We finally parted ways believing that we had an agreement in principle. Several weeks latter, Gerhard Bertcher was appointed CEO of SuSE. A team was sent to Utah and negotiations were begun again in earnest. The plan was to announce at Linux World. Unfortunately, some of the very same issues we had thought we had agreed upon in Spain resurfaced. Caldera finally terminated the discussion and sent the SuSE team home to think over the issues. Gerhard called and we were able to begin again, unfortunately, we could not make the Linux World meetings. I flew down to Houston and had a meeting with Rick Becker and was able to brief him on the discussions we were having. Compaq was very supportive of the idea. However, we did not make a lot of progress. We called Bob Butler of IBM to see if he could step in. IBM had actually invested in several of the other participants and had a little more influence. Bob helped coordinate a meeting in Atlanta. We made tremendous progress. All parties felt that we had created a bond of mutual respect. After the meeting, progress slowed. We called another face to face in New York. Again, we made tremendous progress and set the framework to move forward. To move progress along, Benoy Tamang from Caldera and Roland Dryfuss from SuSE took the lead in the negotiations. Turbo and Conectiva relegated the heavy lifting to Benoy and Roland. We had targeted the first of May to make the announcement.
The first of May came and went. We decided to call a negotiation meeting in Europe. Benoy and Harrison Colter, Caldera’s attorney flew to meet with Roland and SuSE’s attorney in London. They worked several days straight and were able to come up with the first draft of a solid complete draft of the document. We began to brief companies in earnest while we continued to work on the contract. In spite of relegating the negotiations to Caldera and SuSE, Turbo came back with a fairly large list of issues just several days before we were to sign. We wanted to have time to sign and then brief Redhat and Mandrake, but the negotiations went clear through the following week. Turbo wanted to work out the final details with SuSE on the non-Intel platforms. We had scheduled May 3 as the target date to announce. We blew past that date. We then set May 22. That day came and went so we set May 30 as the date. At the end of the week of May 22, we did not have sign able contracts, but we all felt that we would be there by the 30th. Finally, on Tuesday afternoon, all the companies had signed. As previously committed, I called Michael Evens at Redhat. I explained that we had taken longer than anticipated to form the union and we knew that they did not have enough time to consider and participate. I promised to follow up in a couple of weeks. On Wednesday morning, I followed up with Mandrake. They seem genuinely interested and we were to follow up. In some ways it did not matter that Redhat joined because both groups will be supporting the Linux Standard Base initiative. Hence, with the formation of UnitedLinux, the Linux community again showed an innovative way to come together. A well know figure in the Linux industry and good friend, Jon “Maddog” Hall put it the best.
“Open Source licensing leads to each vendor having the same functionality in the base system over time. This effort shows forethought in making that effort a planned, cooperative event. Rather than spend their money re-engineering the basic underlying functionalities of reliability, scalability and availability, these vendors will be able to build a common platform with which to innovate new features that their customer’s desire. I wish the old Unix vendors had learned that lesson.
For many years the analysts have been saying that the marketplace could not support so many distributions of Linux, hinting that only one or two might survive. Once again the Linux community has come back with a unique answer that may prove the analysts both right and wrong. This bold step should give a strong base distribution that will satisfy the needs of the software vendor for consistency and the hardware vendor for support, yet allow differentiation at the upper levels to meet the needs of diverse customers.”
— Jon “maddog” Hall, Executive Director of Linux International.
What Maddog points out is that Linux and in particular UnitedLinux facilitated a new dimension of collaboration even among competitors that previously has not existed broadly in the industry before. This level of collaboration and reliance and trust in a community effort for even competitors is an important mindset and principal that can benefit companies as they deal with the almost overwhelming competitive environment of a global economy, but it is important to understand the history first.
On Monday of announcement week, I flew to Europe. Tuesday, I spoke with Redhat. Wednesday morning I spoke with Mandrake. On Wednesday morning, we held a conference call with the industry analysts. The call went very well. On Thursday we made the announcement. The call and support was significant. The support for UnitedLinux was genuine and evident.
With the formation of United Linux Software, LLC, the Linux industry had reached another major step towards making Linux an industry alternative OS platform. UnitedLinux could have provided a true reference implementation of the LSB specification and a collaborative business model that can sustain ongoing investment in the development of a robust business offering. UnitedLinux could have also provided industry players a direct way of influencing the direction of the Linux technology.
For Caldera, UnitedLinux helped solve several very pressing concerns. We could now get ISV and IHV support for OpenLinux derived from the UnitedLinux base. We could focus our very scarce resources on optimizing that Linux for the retail and distributed enterprise market. We felt we could also infuse UNIX technologies into UnitedLinux to ensure broader acceptance by the UNIX ISV and IHV community including utilities, APIs, and libraries. The SCO management had begun to experiment with a Linux personality on the UNIX kernel and we felt Caldera could now focus on enhancing this layer. I had proposed to the management team that we should contribute the OpenUnix or UnixWare kernel to the open source community. Caldera’s management team felt that by contributing the Unix kernel to open source, we could also contribute significant technology back to the open source community. In turn, the community could help develop the Linux personality and common API layer. Caldera intended to use the kernel and other Unix technologies to enhance the UnitedLinux distribution and create a premiere version. By so doing, Caldera could facilitate a Linux platform for ISVs and IHVs that could literally scale from hundreds of clustered PCs with the Linux Kernel to 36 SMP processors on the OpenUnix kernel using the same Linux APIs. With a common API layer, these sister kernels could have enabled the Intel hardware platform to accomplish almost any application on high end hardware systems.
One of the key benefits of UnitedLinux however, was the precedent setting collaboration between key industry competitors. This mindset was very uncommon in the industry at the time. Some could say that UnitedLinux was unsuccessful as the lack of collaboration from SCO proved to be part of its undoing. However, these corporations did collaborate on a united platform that ISVs and IHVs choose to support. When Novell purchased SuSE Linux, they inherited the collaborative work of the four companies that were fierce competitors. The winner was that the software and hardware developers had just one platform to certify to rather than four. The irony of Caldera and then SCO being both the key catalyst and facilitator of both the creation and disillusionment of UnitedLinux should not go unnoticed.
Fortunately, the principals or mindset which created UnitedLinux have lived on and flourished in the industry. This mindset is almost incomprehensible for many who may read this book. Why would competitors share such intimate information or be dependent upon them for such key technology? A new book entitled “The Starfish and the Spider” details the formation of many leaderless organizations since the Internet and Linux. The book also points out how leaderless organizations cannot be defeated by centralized organizations easily. While doing so, however, it shines further light upon the benefit and power of the trusting previously unheard of intellectual property into the hands of the community and competitors to derive tremendous benefit. They include the story of a Gold Mining Company which was struggling to find new sources of ore on its property. The CEO decided to publish all the highly important and often thought key proprietary information to the community and potential competitors. Within weeks of publishing the information, creative ideas began to flood into the company. They were able to find new ore and within months became one of the most successful mining companies in the world. This new mindset of trust and openness is a vitally important principal for success in this new global and highly competitive world. At a time when most would want to close up and protect their intellectual property, those who will be the leaders and most successful in these changing times will be those who can embrace this mindset.