Off the Beat: Bruce Byfield's Blog

Last weekend, I was exploring GuixSD, the distribution that introduces a universal package manager developed by The GNU Project. Part of the novelty was the lack of a graphical installer, a luxury that most Linux users expect today, but was once controversial, as well as Linux's first encounter with user experience.

Early Linux installations were generally script-driven. Many had the advantage of optimizing installations for the hardware they used, but none were for the casual or curious. Aside from the fact that many system drivers were lacking, the installers were designed for the experienced. Most people who attempted a Linux install usually took several tries, assuming that they didn't abandon the task.

Then the Dot Com era arrived a few months ahead of the new millennium, and Linux became a potentially profitable product. As strange as it seems from a modern perspective, many companies believed that all they required for success was an installer that almost anyone could use.

Others were less sure. Some pointed out that a graphical installer would probably not be optimized for its system, and would probably rob users of the power of choice by imposing defaults. Others suggested that installing the hardware had the advantage of giving users a thorough knowledge of the operating system. Still others waited to see some actual examples, but everyone was waiting to see what would happen.

According to my memory, the first to release with a graphical installer was Caldera Systems. A few years later, Caldera was reincarnated as The SCO Group, gaining infamy for a serious of legal claims over the control of Linux, but, at the time, Caldera was one of the top two or three commercial distributions.

At the time, I was product manager at Stormix Technologies, one of the backrunners in the race to a graphical installer. When a couriered copy of Caldera arrived in the office, the entire staff gathered around a single terminal to watch the process.

As I recall, the installer was not particularly impressive, aside from the fact that it offered a game of Tetris for entertainment during installation, and was advertised -- not quite accurately -- as requiring no special expertise to install. All the same, I watched the Caldera installer for the first time from the edge of my seat. Like many at the time, I was convinced that the first commercial Linux to ship with a graphical installer would immediately become the market leader, and I was all too aware that Stormix's installer was still several months from completion.

After witnessing a second or third Caldera installation, I was less concerned. However, Corel Linux's installer was an altogether different matter. Arriving a few weeks later, its installer was far slicker than Caldera's and several weeks ahead of Stormix's. I consoled myself with the fact that we had a better balance between ease of use and personal choice. We might not be the first to market, but at least our graphical installer was more flexible, I argued, overlooking -- to the great increase in my peace of mind -- that, while Corel was hardly in the league of Sun Microsystems, it was still far more experience and had better marketing connections than our twenty person startup.

However, as you might expect, in six months, having a graphical installer no longer mattered, because all of the top six commercial distributions (among which I inaccurately included ours) had a graphical installer of one sort or another, and no one bothered with novelties like Caldera's Tetris or Stormix's thunderclap at the login screen. Any commercial advantage from a graphical installer was short-lived, and a matter of no more than a few thousand shrink-wrapped boxes.

Thinking about users
If anything, the importance of the race to a graphical installer was technical rather than commercial. For the first time, distributions conceived the idea that their users were not necessarily experts. Instead of being hobbyists who secretly liked to tell war stories of their installation experience, a small but growing number of their users might prefer an easy installation. In fact, they seemed willing to settle for less than optimization if they could get their systems running with a minimum of fuss.

In the first years of the millennium, this user profile was new and different. In building installers, developers started to think more in terms of intelligent defaults rather than absolute optimization, and of window designs that could accommodate both the experts and the newcomers who were starting to investigate free software. When added to the desktop environments that had started to be developed shortly before -- but were mostly too busy catching up with proprietary rivals to have design goals of their own -- the installers helped to create a new sense of priorities.

Even Debian, one of the last holdouts for a graphical installer, improved its text installer from top to bottom. A few years later, when Debian started offering its own graphical installer, the main difference between the text-based and graphical installers quickly became the window decorations.

A similar shift in emphasis, of course, would happen a few years later on the desktop. However, because installers were considered a high priorities, they became one of the places where the practical outcomes of designing for newcomers were first worked out before spreading to the rest of the system. Developers even learned to include some of the optimization that had been the point of the original user-unfriendly installers.

If you are the sort of person who likes to tweak their systems to perfection, you can still get that old-time optimization if you are willing to spend the time. But long ago, the designers of installers have learned to balance optimization and ease of use that is good enough for most us, if not necessarily ideal.