The GNOME project is announced! We’re going to see what the Linux world looked like, and why the GNOME project was started; what was available at the time, for Unix and Linux users, and the beginning of complex desktop environments on Linux, ending on the first few months of the project
It is a long running joke in the GNOME community, that the acronym, or, more accurately, the backronym, that serves as the name of the project does not apply any more.
The acronym, though, has been there since the very first announcement of the project: GNOME, the GNU Network Object Model Environment.
The history of GNOME begins on August 15th, 1997.
NASA landed the Pathfinder on Mars just the month before.
Diana, Princess of Wales, would tragically die at the end of the month in Paris.
The number one song in the US charts is “I’ll be missing you”, by Puff Daddy.
On the 15th of August, Miguel de Icaza, a Mexican university student, announced the creation of the GNOME project, an attempt at creating “a free and complete set of applications and desktop tools, similar to CDE and KDE but based entirely on free software”.
To understand each part of that sentence, I’m afraid we will have to go back to a time forgotten by the laws of gods and men: the ‘80s.
In 1984, Richard Stallman started the GNU project. Don’t bother try to expand the acronym, it’s one of those nerdy things for which the explanation is just not as clever when said it out loud as it feels in one’s head. Incidentally, GNU is the reason why the G in GNOME is not silent. The history of GNU is an interesting topic, but we’ll avoid covering it here; if you want to, you can read all about it on the Free Software Foundation’s website.
GNU was, and it still is, an attempt at creating a Unix-compatible system that is completely free as in “software freedom”; the freedom in question is actually four different freedoms:
- the freedom to use this operating system on whatever machine you want
- the freedom to study it, down to its source code
- the freedom to share it with others, without going through a single vendor
- the freedom to modify it, if it does not do something you want
These four freedoms are enshrined in the “Copyleft” movement, which uses distribution licenses such as the GNU General Public License and Lesser General Public License, to create software programs and libraries, that respect those four freedoms.
GNU is a collection of tools, like the compiler suite and Unix-like command line utilities, in search of a kernel capable of running them — the attempt at creating a working, main stream GNU kernel is, shall we say, still in progress to this day. In 1991, though, Linus Torvalds, a student from Finland, created the Linux kernel, which was quickly adopted as the major platform for GNU, and the rest, as they say, is history — though, like GNU’s history, also not the one you’re going to hear in this podcast.
While the development tools and console utilities were mostly getting taken care of, the GUI landscape on Linux was composed by an heterogeneous set of tools, typically starting with a window manager; some task management tool, like a list of running programs and a way to switch between them; and smaller utilities, like launchers for common applications or monitoring tools for local and network resources.
Each environment was typically built from the ground up and customised within an inch of its life, a system tailored to the levels of micro-management and pain tolerance of each Linux user, and in those days, the levels of both were considerably high.
Large, integrated desktops were the remit of commercial Unix systems, like SunOS, AIX, HP/UX. One of those environments was the Common Desktop Environment, or CDE.
Created in 1993 by the Open Group, an industry consortium founded by the likes IBM, Sun, and HP, CDE was built around Motif, a commercial widget toolkit written in the late 80s by HP and DEC for the X display server, the mainstream graphics architecture on Unix and other Unix-compatible operating systems.
The Open Group quickly standardised CDE, and until the early 2000s, when Linux had started eating most of their lunches, it was considered the de facto standard desktop environment for commercial Unix platforms.
One of the things that Linux and the commercial Unix systems running on the Gibsons had in common was the X graphic system. This meant that you could write applications on your Unix system at university, or at work, and then run it on Linux at home, and vice versa.
As it’s often the case, Linux users wanted to bring some of the tools used on the Big Irons to their little platform, and in 1996 a group of C++ developers led by Matthias Ettrich, created the KDE project, a desktop environment in the same vein as the CDE project, as the name implies. Since Motif was written in C and released under an expensive proprietary license, they used a different widget toolkit, written in C++, as the basis for the desktop called “Qt” (pronounced “cute”), made by a Norwegian company called Trolltech.
Qt was released under a license called “The Qt Free Edition License”, which limited the redistributability of the code: you could get the source of the toolkit, but you could not redistribute modified versions of it. While this was good enough if you wanted to download and build the source on your personal computer, it put some strain on the people distributing Linux and its software, and it went against the Copyleft ethos of the GNU operating system that was the basis of Linux distributions — to the point that an effort to reimplement a Qt-compatible widget toolkit and releasing it under a Free Software license was started by the GNU project, called “Project Harmony”.
In the meantime, Motif was proving to be a hurdle for other free software projects as well.
In 1996, Spencer Kimball and Peter Mattis, two students of the University of California at Berkeley, wrote a raster graphics image editing tool using the C programming language, as part of a university project, and called it the “GNU image manipulation program”, or GIMP, as many have come to regret when searching the name on the Internet. As university students, they had access to Motif for free, so the initial implementation of the GIMP used that toolkit, but redistribution to the world outside university, as well as technical issues with Motif, led them to write an ad hoc GUI widget library for their application, called “The GNU Image Manipulation Program Toolkit”, or “GTK” for short.
A community of software developers, including people that would be influential to GNOME like Federico Mena and Owen Taylor, started to coalesce around GIMP; they saw the value of an independent, free software widget toolkit, so GTK was split off from the main GIMP code repository in the early 1997, and began a new life as an independent project, released under the terms of the GNU Library General Public License. The licensing terms of GTK allowed it to follow the four software freedoms — use, study, share, and modify — but it also allowed the creation of less-free software on top; as long as you distributed any eventual changes you did to GTK under the same license, your application’s code could be released under any other license you wanted. This major distinction with Motif and Qt pushed the newer, volunteer-driven GTK forward, while it filled the gaps with the older, commercially supported toolkits.
GTK had a fairly lean API, and its use of C quickly allowed developers to write “bindings”, that let other programmng languages use the underlying C API with a minimal translation layer to pass values around. Soon after, programmers of Perl, Python, C++, and Guile, an implementation of a dialect of the LISP programming language called Scheme, could use GTK to write plugins for GIMP, or complete, stand alone applications. Compared to KDE’s choice of Qt, which only supported C++ and Python, it was a clear advantage, as it exposed GTK to different ecosystems.
What was GNOME like, back in late 1997/early 1998?
The answer to that question is: an heterogeneous collection of tools, mostly sharing dependencies, and developed together, that occasionally got released and even more rarely did build out of the box without resorting to using random snapshots out of the source revision control.
You had a panel, with launchers for common applications, and with a list of running programs. There were the beginnings of a set of core applications: a help browser; a file manager; a suite of small utilities, mostly GUI ports of command line tools; games; an image viewer; a web browser based on an embeddable HTML renderer; a text editor. Notably, and unlike KDE with its own KWM, not a window manager.
The question of what kind of window manager should be part of a GNOME environment was punted to the users — it’s actually the first instance of a controversial thread on the GNOME mailing list, with multiple calls for an “officially sanctioned” window manager, typically opposed by people happy to let everyone use whatever they wanted — the externally developed Enlightenment was the most common choice at the time, but you could literally run GNOME with WindowMaker; GNUStep; or XF Window Manager 2; and you could still call it “GNOME”.
From a code organization perspective, GNOME started off as a single blob, which got progressively spun into separate components:
- gnome-libs, a series of core libraries based off of GTK
- gnome-core, which contained the session manager and the panel
- gnome-graphics, which contained the Electric Eyes image viewer
- gnome-games, a mixed bag of simple games
- gnome-utils, a mixed bag of GUI utilities, like gtop
- gnome-admin, which contained an SNMP monitoring tool and Gulp, the line printer configuration utility
While GNOME at the time was still a loosely connected set of components, the overall direction of the design was far more grandiose, though. If you remember the GNOME acronym from the announcement, it mentioned a “network object model”.
But what is an “object model”?
One of the things that Microsoft and Apple did with much fanfare, back in the early ‘90s, was introducing the concept of embeddable components provided by both the OS and applications.
You could create a spreadsheet, then take a section of it, embed it into your word processor document, and edit the table from within the word processor itself, instead of flip flopping between the two applications on the limited screen real estate available. The idea was that documents and applications would just be built out of blocks of data and controls, shared across the whole operating system. Those components were available to third party developers in programming suites like Visual Basic, Visual C++, or Delphi, and you could quickly create a well integrated application out of them.
Sun tried to push Java as the foundation for this design; Apple called their short-lived implementation of this technology “OpenDoc”; Microsoft called its much more successful version “OLE”, or: Object Linking and Embedding; and, of course, any self-respecting desktop environment competing with Windows needed something similar to match features.
In order to implement an infrastructure of objects and remote procedure calls that could be invoked on those objects you needed a communication system; OLE used COM, the Common Object Model; GNOME decided to use CORBA, or the Common Object Request Broker Architecture, which was not only meant to be used on local systems but it also worked on the network. As a CORBA implementation, GNOME started by using MICO, a C++ library, and then replaced it with to ORBit, a C library written specifically for the project and addressing some of the MICO shortcomings.
While ideally GNOME would provide components for everything, the initial beneficiary of this architecture was the only recognisable bit of the desktop environment that was visible to the user all the time: the panel.
The contents of the panel were small, self-contained applications that would use CORBA as a mechanism to negotiate being embedded into the panel own window to display their state, or pop up things like menus and other windows on user request.
The core applications started getting CORBA-based components, but we’ll have to wait until after GNOME 1.0 to get to a widespread adoption of this architecture.
Between August 1997 and May 1998, GNOME released various 0.10 snapshots to mark the progress of the project. By June 1998, GNOME 0.20 was released as a “pseudo-beta”, followed, in September 1998, by 0.30, the first named release of GNOME, called “Bouncing Bonobo”.
Between 0.20 and 0.30, though, something had happened: Red Hat, a Linux distribution vendor, founded the Red Hat Advanced Development Labs, hired a bunch of software developers that happened to contribute to GNOME, amongst other things, and the benevolent corporate overlords started taking notice of this Linux desktop.
Nobody really knew it at the time, but the First Desktop Wars had begun.
- Subscribe to this podcast
- Music: “Looking Back“, by Lee Rosevere - CC by-4.0
- Logo: The History of GNOME, by Jakub Steiner
- This podcast is released under a Creative Commons license
- Table of contents.
- The GNOME announcement email
- The GNU Manifesto
- GNU Image Manipulation Program
- Official GNOME window manager thread
- GNOME 0.20
- GNOME 0.30
history of gnome gnome podcast