Off the Beat: Bruce Byfield's Blog

I spent part of the last week reviewing GNOME 2.30. As I worked, I kept returning to the fact that 2.30 is probably the last release of the GNOME 2.0 series, which began in June 2002. That, in turn, got me flipping through the Progeny Debian User's Guide that I did in May 2001 (the last major manual that I wrote), and thinking of all the developments that the last decade or so have seen in GNOME in particular and the free desktop in general.

Writing in April 2010, the state of the free desktop in the early years of the millennium seems unbelievably primitive. The desirability of a free desktop was understood by some, and both GNOME and KDE had been up and running for several years. Yet there were many stone geeks and old timers for whom even a basic window manager was questionable, and the idea of an actual desktop was an abomination.

The free desktop was not helped, either, by the fact that the licensing issues surrounding Qt, KDE's toolkit, were not resolved until September 2000. For a time, Debian seriously debated whether KDE should even be included in the distribution.

Yet even among those building the desktop, the basic concept was crude. The GNOME Human Interface Guidelines were still a couple of years in the future. For many developers, the idea was simply to dump all the command line options into an interface in the quickest way possible. I remember one GUI for Apache that consisted of a couple of dozen tabs crammed with options, and another one for cron whose combo box for selecting the day of the month was so long that at low resolutions it could not be displayed on the screen. In the same way, CD burners existed, but the options they gave were frighteningly complete and obscure.

Things were usually better in GNOME, but releases 1.2, 1.4, and 2.0 could have much the same problems. Unlike today's, the menu had most -- maybe all -- the installed applications listed. Some dialogs, like the one for MIME association, could only make sense to those who had enough knowledge not to need the dialog in the first place. The exit button looked as though it should be on the door of the outhouse, and, in the first user testing by Sun Microsystems, newbies missed it entirely.

As for the references in the interface to the window manager that GNOME was using then, one tester sounded plaintive even on paper when they asked, "What's a Sawfish window manager?" Nothing in early versions of GNOME would provide an answer, let alone explain why users should have to know. With the best will in the world, developers in the first years of the millennium had a hard time comprehending that they were not designing for users at the same level of knowledge as themselves.

Running behind

The general consensus back then was that the free desktop was vastly inferior to its proprietary counterparts, and mostly it was true. In GNOME's case, part of the problem was the limitations of the GTK+ toolkit, which has improved significantly with each release in the last decade.

However, the real problem was that infrastructure took time to develop. Despite some tentative experiments, for the most part, you descended to the command line or edited a configuration file if you had to administration work. The most you could generally hope for was a text-based tool like Debian's dselect, which had been perfectly adequate when the number of packages were smaller, but which by 200 was inadequate for the information it had to handle.

The sole exception was installers, which had started going graphical in 1999. Yet by today's standards, these were applications for the reasonably expert users. Often, the most the newcomer could hope for was that leaving the defaults unchanged would result in a bootable system. Mostly, the tactic worked, but not always.

Yet, even if you confined yourself to productivity, the options were limited. Abiword and KOffice existed, but only in their early stages. The OpenOffice.org project was started in 2000, but 1.0 was not released until May 2002. Firefox soon followed, although usable betas had existed for several years previously.

If you were a developer, you could use the desktop (although you would probably choose not to), but non-technical users had to follow a few well-defined paths if they wanted to be productive. They could not even mount a CD unless they had the root password, because desktops were following Unix security models, and automounting either did not exist or was unreliable.

Forward and Back

The evolution away from these origins has sometimes seemed slow. Often, in looking at a GNOME release, I have been disappointed in the apparent lack of progress. Yet looking back over the 2.0 series as a whole, I now suspect that part of that perception was impatience on my part.

I do not agree with all the changes in GNOME or the free desktop. I sometimes think that convenience has been allowed to triumph over security too often, and that usability risks becoming a synonym for "dumbed-down."

Yet, on the whole, the accomplishments of GNOME and other incarnations of the free desktop have been significant in the past decade. From unpromising beginnings, in which everything remained to be done, today the free desktop has caught up with Windows and OSX -- or close enough that it has started to innovate more often than it follows.

That's something to remember while we look forward to GNOME 3.0 six months from now. GNOME 3.0 marks a new chapter in the free desktop. It is going to be attracting increasing attention, both from those who enthuse over it and those who condemn it as misguided or new. Yet in the excitement of GNOME 3.0, I think it worth looking back at the GNOME 2.0, and congratulating all involved on an impressive work in progress.