Interface problems with KDE, GNOME and Ubuntu suggest that the Linux desktop may be headed in the wrong direction.
The recent history of new interfaces on the free desktop is not a happy one. Three years ago, the release of KDE 4.0 resulted in a user revolt whose like had never been seen. This year, the releases of GNOME 3 and Ubuntu's Unity have produced similar revolts, but on a smaller scale.
Looking at these reactions, I am starting to wonder: Are the developers of free desktops become obsessed with design principles at the expense of what users actually want?
Not too many years ago, the question would have been absurd. Built on the free and open source software (FOSS) principle that developers worked on whatever interested them, the free desktop used to be a hodgepodge in which just getting a dialog to have the same name when it was opened from different places was a major achievement. Back then, the idea that too much attention could be paid to usability would have been unthinkable.
But, beginning with Sun Microsystem's detailed user study of the GNOME desktop in 2001, free desktop developers have been becoming increasingly aware of usability issues. Almost always, the result has been interfaces that are easier to use, more efficient, and more aesthetic.
Nor am I completely convinced that KDE 4.x, GNOME 3, and Unity are exceptions to this steady improvement. However, at the same time, as someone who has grappled with usability issues many times in his career, I have watched the newfound interest in interface design with a growing concern.
At times, the free desktop developers have reminded me uncomfortably of other developers I have known who have just discovered usability and design. Too often, they interpret the tentative conclusions of usability studies as being as incontestable as the laws of physics. From this position, the step is a small one to acting as though designs that attempt to keep these conclusions in mind are as incontestable as the studies themselves.
The reality, however, is more complex. Just as a paragraph can be grammatically correct while its meaning is nonsensical or wrong, so a computer interface can be based on the collected works of Edward Tufte and still contain structures that are awkward and inconvenient.
The truth is, usability principles are always filtered through the basic assumptions of those applying them. If those assumptions are flawed, then the resulting interface will be, too -- especially if the developers ignore the context of what users are accustomed to or believe they want. And, unfortunately, when you are newly steeped in usability, ignoring this context is all too easy to do. Instead, usability experts risk getting lost in the minutiae and arcana of their field
That is what I worry might have happened to some extent on the modern free desktop. Supported by the authoritative tone of usability and design studies, could developers have assumed a false sense of objectivity that has caused them to ignore common sense? At times, the possibility seems all too likely.
Few FOSS projects are as concerned with usability as GNOME and Ubuntu. For GNOME, the Sun usability study proved a turning point, especially when its lessons were codified and expanded into the GNOME Human Interface Guidelines.
Similarly, usability has been a driving concern in Ubuntu from the start. The subject is a particular interest of Mark Shuttleworth, its founder, who stepped down as CEO in 2009 to focus chiefly on design issues. "Ive become very passionate about design and quality, and want to spend more time figuring out how we harness the collaborative process to build better, more insightful products."
Since September 2008, when Shuttleworth first announced usability as a priority, the majority of his blog posts have been on the subject of usability. Regardless of whether you agree with them, these posts have done FOSS as a whole a service by placing usability on the general agenda.
In addition, Shuttleworth and his Design Team have done enough studies of user screen shots to codify at least a couple of important modern design principles: users typically have 3-10 favorite apps, and, in this age of wide screens, "vertical space is more precious than horizontal space."
Yet, at the same time, Ubuntu shows every sign of being dominated by design concerns to such an extent that they sometimes seem to be a detriment.
I have no wish to psychoanalyze Shuttleworth, (whom I have never met, although I respect his vision and effort), but his entries about such subjects as notifications and Ubuntu's latest theme read very much like those of an autodidact happily submerging himself in his latest obsession, to the point where he risks getting lost in the details. Not that such matters are irrelevant, but a corporate leader choosing to present himself to the public chiefly on such subjects does seem more idiosyncratic than practical.
The same can be said about Ubuntu as a whole. If you browse through the archives of the Ubuntu development list, you can hardly miss how the introduction of the Design Team upset the decision-making in the Ubuntu community. Suddenly, all those who had discussed design direction were expected to defer to a set of newcomers unfamiliar with the community's norms.
The situation has since improved, although other design fiats, such as the movement of the title bar button, or the decision to make Unity the default desktop have been caused some upsets as well. At Ubuntu, those making design decisions seem to have veto power over almost everything. To the skeptical eye, this veto can look surprisingly like a claim of objectivity based on a familiarity with usability theory.
The same is even truer of GNOME 3. When you read the design history of GNOME 3, what is striking is the degree to which "from the start, GNOME Shell was a design-led project."
As with Unity, at times this design-centered approach has its points. As the GNOME 3 design history suggests, keeping track of open windows is a problem on all modern desktops, and virtual workspaces, while useful, need to be implemented so that they are easy to use.
However, the same illusion of objectivity that seems to have influenced Ubuntu also seems to have influenced GNOME 3. Design lead William Jon McCann is quoted as saying, "The first thing I did when I started working on the GNOME Shell project was read. I spent a month doing nothing but reviewing usability research into desktop computing. This research reinforced a pre-existing concern: that computers users are increasingly suffering from distraction."
What is interesting here is that McCann reveals the subjectivity of what was to become one of GNOME 3's main design principles. Already convinced that distraction was an issue, McCann found -- unsurprisingly -- the external evidence to support exactly what he was inclined to believe.
Later, when GNOME 3 was released, "distraction-free computing" became one of the benefits emphasized by the marketing team. At the same time, an FAQ answered the question "how do you know it's better?" by emphasizing -- among other things -- "empirical usability research" and "stock usability principles and knowledge." In this way, a subjective impression appears to have been elevated to unquestionable fact.
In neither Ubuntu's nor GNOME 3's case is this transformation of subjectivity into objectivity likely to have been conscious or malicious. But it occurs frequently enough, as people attempt to master the tremendous body of usability theory and are influenced by the apparent objectivity of the scientific and pseudo-scientific voice.
The Forgotten Context
But the danger of studying too much usability theory is not just the temptation to assume infallibility. What is equally important is that it often discourages a close examination of basic assumptions and whether they are complete or can be mapped on to the real-world situation.
My worry is that this outlook has infected both Unity and GNOME. Having identified actual problems, both Unity and GNOME 3 sometimes appear to have moved to solve them without considering whether other problems were equally important, or what new problems might arise from the solutions.
For example, in trying to make the launching of applications easier and freer of error, both eliminated the classic main menu in favor of displays that occupy the entire desktop. This arrangement does improve the launching of applications -- but it does so at the cost of obscuring the windows that are already open and requiring far more clicks and movements away from the active window than the main menu ever did.
The same is true of the decision to design desktops that can be used for mobile devices and touch screens as well as workstations. From the perspective of software engineering, this design principle is sensible because it saves the effort of maintaining at least partially separate code bases.
Yet, from the perspective of a user on a workstation or a laptop, the principle becomes a frustration. Admittedly, a desktop for a smaller screen works better on a larger screen than one intended for a larger screen works on a smaller. Yet it is also true that, at the same time, what works on a small screen can seem cramped and clumsy on a large screen. Nor is a small screen design likely to be equipped to make use of the extra space on a larger screen.
Yet another way that focusing on one design principle can cause new problems is designing with the mythical new user in mind. By "mythical," I do not mean that such a person does not exist, but that they are more of an abstraction than an actual demographic. For one thing, right now, few newcomers to Linux are likely to be completely ignorant of the desktop metaphor. Conversely, if the mythical newcomer does exist, elements such as GNOME's overview screen or Unity's launcher, with its mixture of favorite and running apps, are even more likely to confuse them than anything in GNOME 2.32.
The main point of invoking this alleged new user --however obliquely -- seems to be to provide another level of rationalization for design decisions, especially ones that remove features that more advanced users prefer, such as customizable panels with applets. Perhaps the simplifications associated with the new user might eventually attract actual first-timers, but the main role of the new user right now appears to be to justify the removal of features that existing users want.
That brings up yet another point: although impartially a feature might be a better design, it might not be so in historical context. For example, placing the main menu in the upper left corner of the desktop may be sensible for English speakers because that is where the eye first falls when reading, but is difficult to do because years of using Windows has conditioned millions to looking for the menu in the lower left. At the very least, you need to allow readers the ability to move the menu where they prefer.
Yet this context is almost totally ignored by both Unity and GNOME 3. The simplification of the panel, the repositioning of the task bar, Unity's title bar buttons on the left, GNOME 3's overview -- all these innovations and more not only change navigation on the desktop, but do so while not allowing any alternatives.
And from the perspective of new believers in usability, why should alternatives be allowed? After all, the innovations must be superior to any alternative, because they are supported by usability data. Never mind if the user prefers another configuration -- as newly taught usability experts, the developers must know what is best for users.
The Arrogance of New Experts
An over-emphasis and over-application of usability principles is not the only reason that users object to new desktops. A large reason is probably a reluctance to change. Possibly, too, what might seem like a user revolt can be just as easily explained by the fact that the free/Linux desktop is growing, and attracting a greater diversity of opinions. In other cases, such as KDE's, inflated expectations explain the initially hostile reactions.
At the same time, mainstream usability studies are in many ways foreign to the tradition of Linux. They are generally based on operating systems where restriction of choice is a norm, and the findings of many studies may not be wholly suitable to platforms with a greater tradition of choice.
In this respect, I cannot help but noticing that KDE remains the major desktop least affected by the arrogance of new experts. Admittedly, some KDE developers do have design expertise, but, so far as I can see, that expertise does not generally form the justification for design directions. Instead, KDE still seems inclined to add features because individual developers would like to use them, and to allow alternatives.
In fact, if anything, KDE's design principles -- including the creation of hardware and software sub-systems -- seem designed to encourage alternatives. The overall effect is not as streamlined as GNOME 3 or Unity, but, as KDE has matured from its 4.0 release, the result has been a desktop with more diversity than the current editions of GNOME or Unity.
No one would advocate a situation in which menu structures were different in each application, or the keystrokes for copying and pasting were different. All the same, it sometimes seems that uncritical acceptance of usability principles in the free desktop might be less rationale than is usually claimed -- and, if not carefully watched, as much an obstacle to the development of the free desktop as a help.
Possibly -- just possibly -- mainstream usability may have less to teach free desktop developers than many people assume.