The problem with GNOME 3

Since I started using Linux I used GNOME, v1.2 in those times. It has always done what I needed, maybe not perfectly, and not fully, but for the most part. GNOME 3 changed all that.

I complained about GNOME 3 since day one, and I discussed with GNOME 3 developers many problems with their rationale about why what they were doing made sense, and I foresaw many of the problems they are now facing.

I even started the GNOME user survey, in an effort to make GNOME developers see the light.

I blogged once before about GNOME 3, but only superficially, and even then GNOME developers didn’t take it well. Some GNOME developers might be tired of my arguments in Google+ and other media, but it’s time to dive into the issues with GNOME 3 in this blog.

Two years after the first release of GNOME 3, once GNOME developers have had the time to polish the rough edges, it’s time to make the call; GNOME 3 sucks. Here’s why.

The purpose of GNOME

Before we start, we need to clarify the purpose of GNOME, or rather; the purpose of any software project in general, which is to be useful to the users. Your software might the most efficient at doing certain task, it might have the simpler code, and the best development practices, but it’s all for naught if it’s not useful to anybody.

Software has to be useful for other people, otherwise nobody will use your software, and nobody will contribute to your software. Quite likely there will always be other software that does something similar to yours, so you need to convince users that your software is better than the alternatives, usually by providing a good user experience.

Once the user has spent time analyzing the alternatives and has chosen your software, the user expects your software to keep working, and in the same way. If the software keeps changing the way it behaves from one version to the next, it’s not achieving it’s main goal; to be useful to the users. If you do this, your users will move on to another project that might not have as many features, but at least they can rely on it and they don’t have to spend any more of their valuable time learning what broke now in the new release.

More important than providing a good user experience, is to not break existing user expectations.

You would think this is as obvious as a leopard’s spots, yet many software projects, including GNOME, don’t seem to understand this fact.

No project is more important than the users of the project. — Linus Torvalds

What users want

The problems I’ll talk about are not my invention, but conformed by GNOME user surveys, which show the most important issues according to the users:

1. Lack of configuration options
2. Developers don’t listen to users
3. Bring back traditional interface

I knew these were important issues myself before the survey, but a survey was needed in order to know for sure. Of course, even after the survey, GNOME developers reject these issues are important to their users.

Developer attitude

In my opinion the main problem with GNOME is not just the code, code can be fixed, but the attitude from the developers (which is reflected in the code), and as users made it clear in the survey.

Many open source projects would kill to have the user-base GNOME had, and welcome their input with open arms, but GNOME neglected their users, they thought they were irrelevant, and they tried to dismiss their complaints with typical defenses, which of course don’t make sense.

People hate new things

By far the most widely used defense for the changes in GNOME 3 is that some people are backwards, and always hate new things, so it’s no surprise that they hate GNOME 3.

Any rational person would see the problem with this claim, but unfortunately I must spell it, because I had to explain it at great length to GNOME developers, and based on the responses, one would think they are incapable of seeing hasty generalization fallacies.

Say 50% of users hate GNOME 3, how many of them have legitimate reasons, and how many hate it just because they always hate new things? It’s impossible to know. So to dismiss every complaint saying “people hate new things” is totally stupid; you cannot know beforehand if they belong to this group or not, you first need to hear their complaints, and then decide if the complaints are invalid or not.

Saying “Jon doesn’t have legitimate complaints” makes sense (after analyzing his complaints), saying “people hate new things” does not, and GNOME developers do the later when people complain about GNOME 3.

Users don’t know what they want

Really? People don’t know if they want a punch in the face or not? Some people might not know what they want, but they definitely know what they don’t want.

Moreover, that’s a very rudimentary notion of user choice. Malcolm Gladwell explains rather well how it’s true that people don’t know what they want, or rather; they can’t tell you what they want, but when they get a taste of it, they most definitely can tell you: this is what I want.

If it wasn’t clear by now; GNOME 3 is not what a lot of people wanted.

People complained about GNOME 2 too

Did this actually happen? If you google GNOME 2 for results between 2002 and 2004 you would find barely (if any) results that show negative comments, so if it did happen, the backlash probably wasn’t that bad, and it probably died rather quickly. Check the thread in Slashdot, and try to find one comment suggesting to go back to GNOME 1; I couldn’t find any.

Did anybody create forks after GNOME 2 was released the same way people forked after GNOME 3? No.

Yes, people complained about GNOME 2, but not nearly as much as they do about GNOME 3. There’s even an entire wikipedia article dedicated to the controversy over GNOME 3.

Sorry GNOME developers, none of your excuses for not listening to your users make sense. You should always listen to your users. Without them your project is nothing.

Configuration

The biggest complaint users had was the lack of configuration options, which could have been solved very easily, if the developers had more than two neurons in their brain, but instead, there’s only sub-par alternatives.

Advanced mode

GNOME 3 got rid of a bunch of useful configurations users relied on, and to defend themselves they said you could use gsettings manually, or install (or write) an extension, and so on, but I argued that there should be an “advanced mode” that the user can optionally turn on in order to further configure the system the way they wanted.

GNOME developers argued that this was a bad idea, because more code would need to be maintained which is absolutely not true for gsettings; the configurations are already there in the code, the only thing missing is the UI. Then they argued that it would mess up the configuration UI, which they wanted to keep clean, but adding a single check box can’t possibly mess with anything. In the end, I never received a single satisfactory answer of why an advanced mode didn’t make sense.

Tweak tool

The tweak tool is the closest thing to a sane configuration interface, however, it should be cleaned up, distributed by default, and integrated in the GNOME control center by adding an “Advanced mode” checkbox that when switched on enables these settings which should be distributed among the different “capplets”.

Even then, the amount of configurations available is not enough.

Shell extensions

Finally, there’s shell extensions which sound like a great idea at first, but if you think about it for thirty seconds, you realize the problem; there’s no APIs for extensions. This means extensions bring many problems.

First, there’s no guarantee that an extension will work the next version of GNOME, because it’s modifying the JavaScript code directly, and there’s no guarantee that the code will remain the same. There’s no extension API like with Chrome, or Firefox. The result is not unexpected; extensions break very easily from one version of GNOME to the next.

The second issue is that extensions can easily break GNOME, because they can modify absolutely anything in the shell.

gsettings

Another option is to manually change the configurations through the command line. This is obviously not user-friendly, and not ideal either.

GNOME developers made the mistake of not allowing configurability directly into GNOME, and as a result they have made the much needed configurations second class citizens, than no matter what you do, they don’t work as expected. The problem is very easy to solve; the configurations should be integrated into GNOME, and activated by a single switch, so their normal configurations are not disturbed. But talking sense to these guys is next to impossible.

Classic mode

Finally, after so much backlash from the users for years, GNOME developers had to implement the classic mode, so users could be able to run their beloved GNOME 2 interface with GNOME 3 technology.

Unfortunately, GNOME classic is more of a hack than anything else; it’s full of bugs and inconsistencies. Sure, this mode is still in its infancy, and will probably improve in the next releases, but it’s safe to bet that it would never be as polished and usable as GNOME 2 was.

Sorry GNOME devs; too little, too late.

Suckage

I am not going to list every single reason why GNOME 3 sucks, I would never finish this post, and there’s plenty of people that have already added their share (I’ll list some of them).

Personally for me the most obvious way to see that GNOME 3 defaults are brain-dead is the way alt+tab works. The whole purpose of work-spaces is to separate work, instead of having all your browser windows, terminals, editors, email client, and so on, on the same work-space, you split them among different ones. Even their own design references for GNOME shell say so. So, to make alt-tab cycle through the windows in all works-paces instead of only the current one is idiotic; it undermines the whole purpose of work-spaces.

Linus Torvalds schools Lennart Poettering on the importance of users

Instead of accepting this obvious fact (which many users complained about), they disregard any feedback, and say “oh, you can use an extension for that”. Never mind the fact that a huge percentage of their users need to install the extension just to have a sane behavior, and that the extension can break at any moment.

Another quite obviously stupid behavior is to switch to an already running “application” instead of opening a new one. For example; if I press the Windows key, and type chromium, I expect a new browser window to open, but no, instead I’m unexpectedly dragged to another work-space where there’s already a browser window. This makes sense on single-window applications, but not on other ones.

Linus Torvalds and others complained about this, and I proposed a solution; add a menu item when right clicking on the application that says “Always open new window”; problem solved. Now each time the user tries to open this application, it’s done in a new window. But no, users don’t know what they want; GNOME developers know better what’s best for us.

The list of stupidities in GNOME 3 is never ending:

• Gnome 3.8 Review… Still Sucks!
• GNOME (et al): Rotting In Threes
• I hate Unity. I hate GNOME. I hate Windows 8. The ultimate desktop search continues.
• GNOME 3: Why It Failed
• A shiny new ornament for your Linux lawn: Ars reviews GNOME 3.0
• The Grumpy Editor’s GNOME 3 experience
• GNOME Shell, extensions, and control
• Four tweaks to bring back missing functionality in GNOME 3.0
• Why GNOME refugees love Xfce
• Yes, Gnome 3 Sucks
• Why Gnome 3′s Fallback mode sucks
• If it ain’t broke, don’t fix it
• Ubuntu has jumped the shark and may soon be bitten by it
• Has Gnome 3 decided that people shouldn’t want screen savers?
• GNOME developer quote of the day
• GNOME 3 – Built for people who would never use Linux in the first place
• Five Linux Desktops That Aren’t Unity or GNOME 3
• Something Gnome3 and Unity could Stand to Learn from Windows 8
• Debian Beckons Ubuntu Refugees to Come Home
• Staring into the abyss

The Linux way

What GNOME should have done is simple; don’t ever, ever break user experience. This is how the Linux project has managed to become the most successful software project in history. There are no forks of the Linux kernel, simply because there’s no need; user experience is never broken; what works in v2.0, works in v3.0, and will work in v4.0. New features are added carefully, making sure nothing breaks, and users always have a way to make the system work the way they expect it to work.

GNOME shell should have been developed alongside GNOME 2 and users should have given the option to turn this mode on and off. Slowly, GNOME shell would get more features from GNOME 2, until it was a full replacement, and at the same time GNOME 2 would evolve to integrate GNOME 3 technologies (like GTK+ 3.0). Eventually GNOME shell becomes the default, but always with the option to use the classic interface. After some cycles, the GNOME shell interface is bound to be a full replacement for the classic one.

Unfortunately, they chose not to do this.

Going this way requires more effort, of course, but it’s the only way to move forward without loosing developers and users along the way. GNOME developers argued that they didn’t have the resources to do such careful move, but that is clearly false, as the developers working on MATE, Cinnamon and GNOME classic show; there’s developers interested in making things work for the traditional mode.

They chose not to listen to the warnings, and they became impatient and went forward with a move that had the potential of angering a lot of users. Well, this potential was fully realized, and now they are paying the consequences.

They were wrong

They argued this was like GNOME 2; the people complaining about it will eventually learn the new ways, and most will be happy. They were wrong.

Even to this day people keep complaining about GNOME 3, how the interface doesn’t make sense, how the developers don’t listen, and how the design is brain-dead.

There’s no other way to put it; GNOME 3 was a mistake.