What went wrong with the KDE 4 release?

When KDE 4.0 was released in January, it was supposed to be the foundation for a new era of desktop development. But as 4.x versions began finding their way into distributions, negative reactions began to obscure other ones. With the upcoming 4.1 release due at the end of this month, it’s hard to avoid wondering: what happened?

To a degree, the answer seems to implicate everybody involved, from KDE and the distributions that ship it to the free software media and users. In doing so, the answer highlights some of the changing relationships in the free and open source software (FOSS) community.

Finding answers matters because the user revolt — or the perception of one — that KDE seems to be facing is almost unprecedented. KDE took a giant technical leap forward in each of its other releases, but the reactions to the 2.0 and 3.0 releases were very different from 4.0’s reception. Like the latest release, KDE 2.0 had early performance problems, and some applications were slow to upgrade, but users seemed largely content to wait for improvements. Similarly, although KDE 3.0 was greeted by accusations that the release was mishandled on the kde-devel mailing list, both user and reviewer reactions were generally positive. Nor has any other FOSS project received a backlash like the one facing KDE.

By contrast to earlier releases, what happened to KDE 4 is harder to make sense of. Despite its misleading version number, KDE 4.0 was repeatedly described as a developer’s release — not for the average user’s desktop. Wade Olson, the press contact for KDE in North America, points out that most major reviews of KDE 4.0 clearly communicated its state.

However, starting in April or May — about the time that the first versions of KDE 4.0 were finding their way into distributions — that message seems to have been lost, and users started to react as if KDE 4.0 were a finished piece of software. Within that frame of reference, some of these reactions were justified, especially those about the loss of features and customization options in the available binaries. Other reactions, such as criticisms about the new menu or desktop manager, may have been indicators of a fear of change as much as legitimate complaints. Some, like the numerous complaints about stability, were as much the responsibility of the distributions as of KDE. Still others, such as the fear that the emphasis on Dolphin as a file manager meant that Konqueror would be slowly phased out, seemed to have little basis in fact. However, regardless of their validity, the negative reactions slowly became dominant, appearing on various KDE mailing lists and as comments on KDE.news and even as repeated thread hijacks in the middle of Groklaw discussions.

These reactions were accompanied by articles in the same vein by professional and semiprofessional writers. At one extreme, Linux Hater’s Blog zeroed in on KDE and core developer Aaron Seigo as continual targets for abuse. At the other extreme, veteran FOSS journalist Steven J. Vaughan-Nichols, after expressing his personal dislike of the KDE 4 desktop, suggested that the project needed a fork because the current developers had lost touch with users. Such writings both echoed user sentiments, and, perhaps, created them.

Having worked on KDE 4 for more than three years, KDE developers reacted with understandable anger. In particular, Troy Unrau, best known for his “Road to KDE 4” articles, went so far as to say in his blog, “KDE and open source is not ever obligated to please users. We are not obligated to fix bugs. We are not obligated to implement things that you demand. We are not obligated to provide open forums for you to attack us personally.”

A week later, Unrau apologized, but his rant had already fuelled the flames. When Unrau put his KDE activities on hold for personal reasons, his departure was widely seen as a reaction to the situation. Many saw Seigo’s suspension of his blog so that he could focus on coding in a similar light, although he himself explains it as a wish to step down after more than a year of being the chief public relations figure in KDE.

Puzzled responses and justifications

In the last couple of weeks, Seigo and other key KDE developers have begun taking a more dispassionate view of the situation as they try to figure how their dream turned into a nightmare, and to prevent a similar situation from happening in the future.

To start with, Sebastian Kügler observes that the negative reactions are not the only responses to KDE 4. He points out that, since KDE 4’s release, 166 new developer accounts have been opened for KDE’s versioning system — almost one every day. He also suggests out that positive reactions tend to be ignored in favor of the more sensationalistic negative ones.

Still, even with these encouraging signs, Kügler says that the general reaction “puzzles me. The complaints feel disconnected to our actual performance.”

Similarly, while Wade Olson admits in his blog that KDE might have missed “some features or level of perceived stability that we predicted,” he also maintains that “we’ve been promoting nothing but honesty from a marketing perspective since inception.”

A wicked problem

A more complex view is expressed by Aaron Seigo, who refers to KDE 4’s reception as what management theory sometimes calls “a wicked problem” — that is, as “not a problem where there’s one cause and one solution.” Instead, he sees the reception as reflecting a number of different tendencies in KDE in particular, and FOSS in general.

Seigo begins by acknowledging that the sweeping changes in both back end technology and desktop functionality may have been more than many users could handle. He suggests that, after years of trying to equal Windows and Mac OS X desktops, KDE has finally succeeded, and is now looking for ways to innovate. The trouble is, this change leaves users without a frame of reference.

“It’s harder for people to have a sense of what we’re doing and what we’re reaching for,” Seigo says. “[KDE 4] is such a huge and new set of ideas that, unless you immerse yourself in it on a daily basis, it’s not that easy to get.”

At the same time, Seigo suggests that the KDE project “did not do as good a job as we should have of managing some of the conversations” on its sites and mailing lists. While KDE is used to unmoderated conversation, Seigo sees that a small number of posters “who really don’t have a lot of care or trust for anybody” were allowed to dominate the discussion, and “the tone was set by this small group of people who were loud, ill-mannered, and obnoxious. And that set a tone, and a sort of mob mentality set in.” As a result, more reasonable voices were lost or stilled.

For Seigo, another part of the problem is that KDE may have misperceived its audience. “Personally, I naively thought that we had really proven ourselves with KDE 3. We had taken the promise of KDE 2 and matured it to KDE 3.5.9. And then we were going to attempt to replicate the results of that previous effort and take it to a whole new level.”

But what Seigo and the rest of the KDE community may have failed to realize is that, with the recent popularity of FOSS, many KDE users lacked this historic sense — and, therefore, did not have the sense of trust he expected.

In addition, he says, “There’s still very much a consumer model in people’s minds and not a participatory one.” In other words, some users, rather than trying to contribute to the project, reacted more as customers who had a right to demand satisfaction, and as though the only way they could get their complaints addressed was by causing a disturbance.

Another contribution to the problem, Seigo suggests, is the varying ways in which distributions handled KDE 4.0. While some, like openSUSE, took time to polish KDE 4 and presented it as an experimental option, others “released packages that simply don’t work. And these days, the distributions are pretty much our public face. So when the distributions don’t get it right, we look bad.”

Not all distributions seem to have communicated — as the KDE project expected they would — users that KDE 4 was experimental. Instead, they seem to have been motivated by a number of factors, such as the fact that support for the Qt 3 tool kit used by KDE 3 would no longer be maintained after July 1. This situation left distributions with the choice of a release that would depend on unsupported libraries for some of its lifecycle or shipping KDE 4.x. Given the competition between modern distros to be the first to include the latest software, many opted for shipping KDE 4.

In many cases, this decision increased the dissatisfaction. On the Fedora list, for example, many users complained long and bitterly about KDE 4’s unfinished state, and some queried what it would take to ensure that KDE 3.5.9 packages would be available in the next release.

Yet another factor is the role of the FOSS media, according to Seigo. Because the negative reactions were so violent and colorful, they may have encouraged journalists attempting to be fair to give them more attention than they deserved. In turn, those wishing to find faults focused on the negative aspects in reporting and ignored the positive, creating a vicious circle of escalating criticism.

Furthermore, Seigo sees a switch in the FOSS media from advocacy journalism to an adversarial position “that doesn’t exist and shouldn’t have to exist.” According to Seigo, the role of the FOSS media is to “interpret the bias of KDE, and they have failed to do so.”

“You guys are part of the machine,” he says. “And you are a big part of whether what goes out is accurately reported or fairly reported. You become part of the success or not.” In other words, in the case of KDE 4, the FOSS media failed to do its job.

However, Seigo reserves his strongest criticism of the media for sites like the Linux Hater’s Blog. “I think it’s exceptionally sad that they have received the level of acceptance they have for what isn’t satire, but is pretty puerile and negative. There’s a difference between satire and just being destructive. Those same things can be said in a way that isn’t destructive.”

Working towards solutions

The KDE project rejects out of hand the idea that technical changes might be slowed so as not to overwhelm users. No doubt, after so many months of planning, admitting mistakes in direction would be difficult, but the issue does not seem to be even considered.

“I really hope that the message that gets out is not that innovation is dangerous and will wreck your project,” Seigo says, addressing participants in other projects. “That scares me to death. We’re at the point where we’re turning a lot of concerns innovation-wise and technology-wise, and I really hope that other projects don’t look at this and say, ‘Wow, if I do something really cool, I’m not only not going to get rewarded, but be beaten up for it.’ Don’t lose your nerve. Continue following your passion, and don’t be afraid to make mistakes.”

Instead, the KDE team is placing much of its faith in communication to solve existing problems. As Summer of Code student Chani Armitage blogged, the question is, “How do we keep communication between developers and other contributors and users open, without people being overwhelmed?”

So far, possible answers to this question include a user advocacy board in KDE, and a glossary and FAQ to accompany new releases. An especially popular suggestion is a code of conduct on KDE sites and mailing lists.

“The code of conduct is a way to let people coming in know what our culture is and what is acceptable and what isn’t,” Seigo says. “It allows us to create a description of what to most of us is common sense, what our community actually stands for, and then is there for everyone to see and reference. And when exceptional steps have to be taken, then the person who is going to take those steps can look at it and go, ‘OK, I’m justified in doing that.”

Most of all, though, KDE is counting heavily on what it delivers in 4.1 to make people forget the problems with 4.0. “This will wash out over time,” Seigo says. “The biggest thing that we can do is executing and delivering. People who look at 4.1 are seeing a very different animal from 4.0, which is what we said at the beginning.”

How much clearer communication will help may take a while to determine. How much a strong release can calm the disturbance will be revealed when the community can assess 4.1 at the end of the month. But the real question — and what seems to worry many KDE developers — is whether the tainted atmosphere around 4.0 will cause people to reject 4.1 without giving it a chance.

It’s a question that members of other projects may want to watch closely. More than anything that the project did or didn’t do on a technical or PR level, the problem seems to have been caused by some unexpected but fundamental changes in the ways that FOSS projects, users, distributions, and the media relate to each other. If that’s the case, other projects, especially large ones, might find themselves facing a similar revolt — possibly for no better reasons than KDE has.