Lou's Views: Converging on a Linux Desktop? Part IDec 18, 2000, 13:30 (22 Talkback[s])
(Other stories by Lou Grinzo)
Is Linux starting to converge on a desktop standard, and if so, is that necessarily a good thing? I've been pondering these issues for quite some time, and some recent market survey information that's come to my attention has shed at least a few stray photons in their general direction. At this time in Linux's lifespan these are both Big Topics, so I'll talk about the "is it happening" issue this time around, and leave "is it a good thing" for the next Lou's Views.
The impetus for this heightened introspection was a posting by Chris Schlaeger of Kamp KDE (carried here on LinuxToday.com), in which he said, "Recent studies show that KDE is used on more the 70% of all Linux desktops. We could fight for those remaining 30% but given that Linux has less than 5% of the overall desktop market we should rather target the 95% of desktop users than compete with our friends from the GNOME project."
Seventy percent sounds mighty impressive, and, more to the point, it would seem to be irrefutable proof that not only does KDE have a commanding lead over GNOME and every other contender, but that the convergence on a single desktop is already well under way, even if no one noticed. Frankly, it seemed a little too good to be true; KDE is a very good desktop environment, no doubt about that, which is why so many people (including your faithful correspondent) use it. But the more I thought about it, the less credible that number seemed in light of the plethora of window managers and desktop environments for Linux. Another factor arguing against the 70% solution is the passion some people still have for opposing KDE and Qt at every turn, thanks the Licensing Issue That Will Never Die. Instincts piqued, I doffed my 1940's-style reporter's hat, complete with PRESS card stuck in the band, and started sending e-mail. I first asked our friends at KDE where that number came from, and they cited the market research company Evans Data Corporation. A quick check with them revealed some interesting, if not completely definitive, details.
First, the number came from a survey of about 300 developers that was published in March of 2000. That's right, developers, not simply people using Linux on a desktop. Second, the respondents were allowed to pick more than one answer, and the complete results, as provided by Evans Data and quoted with their permission, were:
Count Percent of Percent of Responses Cases KDE 217 25.7 70.7 GNOME 196 23.2 63.8 FVWM 120 14.2 39.1 AfterStep 69 8.2 22.5 Windows Maker 62 7.3 20.2 Motif 43 5.1 14 Lesstif 42 5 13.7 CDE 40 4.7 13 None 29 3.4 9.4 Other 28 3.3 9.1
Obviously, the first lesson is not to jump to conclusions, and to read any quoted statistics very carefully. Chris Schlaeger never said they had a lock on 70% of the desktop market, which is what I, and no doubt many other people assumed he meant. He merely said that KDE was used on 70% of desktops. You can beat him up for saying this percentage applied to "all Linux desktops", however, since that's clearly not what the survey measured. (The kind people at Evans Data Corporation also made it clear to me that no one should quote that number without making it clear it was derived from a question allowing multiple responses. All things considered, I have to agree with them.)
But wait! as they say on infomercials, there's more! Evans Data Corporation also mentioned that there was another study they published in September of 2000, six months after the one Schlaeger quoted, that asked the same question. This time, the results, exactly were:
Desktop Environments Count Percent of Percent of Responses Cases GNOME 185 26 66.3 KDE 177 24.9 63.4 CDE 38 5.3 13.6 None 35 4.9 12.5 Other 21 2.9 7.5 Desktop Windows Managers Count Percent of Percent of Responses Cases FVWM 72 10.1 25.8 Windows Maker 70 9.8 25.1 AfterStep 52 7.3 18.6
Aside from the obvious facts that GNOME passed KDE for first place and that they're the dominant desktop software, what can we glean from comparing the March and September snapshots about the desktop situation in general? Not a lot, in my opinion, since only developers were questioned, not "real" users. I've long believed in the "killer critical mass" theory that says developers often lead the market, and that the winning platform in terms of market share is the one that does the best job of capturing developer time and mindset. (Exhibit A: Windows' 90% share of the desktop market. Say all you want about Microsoft's business practices, it's true that they did a masterful job of locking in developers early in Windows' lifespan, and that doing so was key to the later marketing maneuvers actually working.)
Lesstif and Motif were both removed as separate answers in the September survey, so presumably people who used those also chose "other" as one of their selections. Yet "other" actually declined from the first survey to the second, which suggests (but doesn't prove) that the desktop choices among developers are narrowing slightly.
Notice that the window managers in both surveys even split--FVWM and AfterStep were down, with Window Maker up significantly. If you're so inclined, you can try to figure out how many desktops the average respondent uses. Throwing out the "none" votes this works out to about 2.7 desktops/person in the march survey, and about 2.3 in the September survey. Given that these still include the "other" category, which is in itself a wild card, since we don't know how many environments any one person is indicating when he or she picks this response, it's hard to say even the developers are beginning converging on one desktop. I think it's clear that all users, my fellow developers as well as real people, will they will eventually succumb to the pressure of the "killer critical mass" effect as more and more KDE- or GNOME-specific apps are created, and both environments continue to improve at a stunning pace.
In my opinion, the jury is still out on whether we're seeing the glaciers converge (although I'm eager to see the March 2001 numbers) . But the question of whether consolidating on a small number of Linux desktop environments, or even just one, is a good thing is something we can talk about much more concretely. Which I'll do next time.