How to Improve Open Source User Interfaces

Published 2008-08-08
This is one of my favorite blog posts

Programmer Michael Thomas revisits an old theme: Why Free Software has poor usability, and how to improve it. This was a theme I heard a few times at OSCON as well. Thomas unfortunately gives no space to respond, but his loss is my gain. (Specifically, I gain a blog post on the theme).

Thomas lists 15 reasons for poor usability in FOSS (Free and Open Source Software) programs, most of which can be boiled down to: “designers don’t contribute to FOSS projects” or “programmers are smarter than interfaces and therefore don’t value good ones.” I think he’s right. I also think there’s an important meta-explanation that he kind of dances around:

Interface design is not much like programming.

If FOSS project owners are serious about improving the usability of their projects, I think they need to focus less on grafting interface design tasks into an ecosystem optimized for programming, and focus more on altering the ecosystem to favor interface design. I think breaking down how interface design is not like programming reveals a few potential courses of action for reforming FOSS ecosystems.

1. Interface design is a Gestalt task.

A truly beautiful interface is seamless and uniform. It requires seeing forests more than trees; it may even require envisioning a total view of the entire experience. There is almost no way to “jump in and help out” on an interface design in a meaningful way. Successful projects with many interface designers typically have a small number of lead designers (usually: n = 1) who delegate tasks to junior designers. The beautiful Gestalt is the result, not of democracy and open collaboration, but of ruthless tastemaking. Good senior interface designers elicit and accept feedback, and work closely with users and programmers in a spirit of collaboration. But in the end, they are willing to unilaterally kill bad ideas and advance good ones. The fundamental nature of good taste is its exclusivity.

My remaining points can all be viewed as corollaries of the theory “interface design is a Gestalt task.”

2. Interface design is not as much fun as programming.

I’ll explore this in more detail in my next three points, but let me dip into anecdote a little here. Of the hats I wear (graphic designer, interface designer, web developer), the only task I prefer to do in my free time is development. I get a happy charge out of starting up MacVim that I just don’t get from starting up Illustrator. I read books on programming when I’m on the crapper. Every designer who works even a little bit with code must, at some point in his or her career, be forcibly kept away from doing yet more code, because it’s so tempting to dink with the code instead of the pixels, even when we’re fucking up the code and making more work for the better developers downstream. Many designers love to spend their free time pushing pixels, of course, but the artsy ones tend to do actual art: painting or woodworking or sculpture or such. The nerdy ones spend their free time writing recursive draw routines in Flash, or making CSS Zen Garden designs. In other words: coding.

Occasionally I’m pressed by a layperson to describe how difficult graphic or interface design is. Usually I fall back on this analogy: “executing a design feels a lot like writing a term paper.” (The bigger the project, the bigger the term paper.) That’s how much fun it is. If you think writing term papers in college was fun, you might like interface design.

3. Interface design offers no “aha” moment

Everyone who has ever written a Hello World program knows what I mean here. We get a little charge when something you built — no matter how trivial — actually works. There is no moment like this for interface design. Interface design is a long process of slow improvement. The design might be better at the end of the day than it was at the beginning, but there was never a moment during the day when you flipped a switch and the UI compiled. The quality of the interface is purely a function of the amount of time you put into it — no shortcuts, no macros, no aha moment, no unit tests to pass, no checksums to verify. Most UI designs are “done” when you hit the deadline to hand it off to the development team.

4. Interface design offers very little glory.

A competent interface design, like the graphic design of the phone book, is invisible. The harder you work at it, the more invisible it becomes. A few interface designs — Panic’s Coda, or Omnigraffle come to mind — have a sparkly beauty or particular twist that really stands out, but most “good” interfaces simply (and properly) fade into the wallpaper.

5. Interface design artifacts aren’t functional.

Some of the best UI designs I’ve ever seen were committed on whiteboards or sketchpads. In the majority of cases, the UI designer produces a Visio document or Photoshop file. The product isn’t functional because there’s nothing to click.

Really tricky designers might use Interface Builder, or Tk, or HTML prototypes, or similar code tools to build a kind of Potemkin Village interface: the taps are all there, but they aren’t connected to any plumbing. You can turn a tap but nothing will come out. Nobody uses the “Interface Design” as anything other than a blueprint to build something else.

A person can gain great satisfaction from blueprinting a garage, I’m sure, but it’s a totally different pleasure than building a garage and then parking your car in it.

6. Interface design tasks scale poorly.

This is not strictly true, as I noted above. Experienced senior UI designers are adept at delegating tasks to junior designers. But the key word here is “delegate”. I can think of few things more counter to the FOSS ethic than the arbitrary elevation of a single person to the status of “senior” anything. But when it comes to good interface design ... there it is. If you break apart UI design tasks without a master vision, the result tends more often to be disjointed and fragmentary — the exact opposite of a usable or beautiful user interface.

The kinds of problems that get solved by UI design are of a different order than those solved by programming. I can write a script that wraps lines in a file with HTML paragraph tags. Then I can add a function to create bullet lists. Then I can add a function to format headers. I can keep adding layers of functionality to my script until it becomes an entire piece of software. From a tiny acorn came HTMLFormatomatic4000, or whatever.

In fact, I would extend this point to read: as a general rule, the more designers who work on a user interface, the worse it will be. Too many cooks really can spoil the soup.

7. Job demand for interface designers is higher.

I don’t mean to imply that interface designers are better compensated (because it’s probably not true). But I do mean that, if I have a few spare cycles, someone is almost always willing to pay me for those cycles. I might find a FOSS project about which I’m passionate enough to donate those cycles, but that poor project is always going to be competing against filthy filthy lucre. For a task I don’t find as much fun as programming. On a project where designers may be treated with active contempt as “decorators” or “icon designers.” And which probably has no resources to buy me, for example, an Adobe Creative Suite license, a new quad core Mac Pro and a 22" monitor.

I have no doubt that FOSS projects fight for developer resources in the same way. I’m sure most developers worth their salt could fill their weekends with freelance work the way most designers I know do. I can’t much speculate why they don’t. I guess I’d theorize that FOSS projects actually do provide a fungible benefit to programmers. If I contribute code to a well-regarded FOSS project, my nameshare rises and my ability to get paid for my expertise on that software would also rise. I’m having trouble seeing how contributing interface designs can provide the same benefits.

How to Improve Free Software Interface Design

I think understanding the nature of UI/usability design as a Gestalt task is key to improving free software interfaces. Make your project’s ecosystem more amenable to seeing and driving the Big Picture, and you’ll attract more interest from A-List designers. And I think you’ll improve your result. I’m still chewing over how to do this, but a few reforms spring to mind.

1. Restrict UI participation in your project

Thomas and others, used to a FOSS ecosystem where More Developers = More Participation = Better Feature, miss the allure and challenge of working on the Big Picture. When the picture is big, the glory is restricted. You can probably name the director of 2001: A Space Odyssey, but can you name the editor? Directors make their careers by not sharing. (When you consider the unfun, unglorious nature of UI design, maybe you’ll gain a little appreciation for why so many designers act like prima donnas. It’s because they are constantly beset with criticism for a task that compares with writing term papers in its pleasurability. Their only nonmonetary compensation is sole glory — best to hog as much of it as possible, by fending off all feedback if necessary. I don’t care to do business this way but I understand why many designers do.)

2. Invite designers to your project

The easiest way to shift a focus to the Big Picture is to invite a select few individual designers to work on your project, and guarantee that they’ll hold the reigns of the interface. If they need help, let them invite their own collaborators (or junior designers). This flies in the face of the Free Software ethic, I know, but remember that while your Interface Guru has veto power over all interface decisions, the community ultimately has veto power over the product. Other designers are welcome to volunteer their efforts, but if your selected designers say they can’t use them, remember the maxim about cooks and soups.

3. Compensate your interface designers

Maybe not with money, but certainly with resources, and definitely with accolades. Prominently feature the contributions of your UI designers on your project website (you do have a project website, don’t you? And no, SourceForge and Google Code don’t count.) Designers, much like developers, attract work through their reputations. What are you doing to massage your designers’ reputations?

4. Assign your designers a programmer au pair

Many excellent designers know (and care) very little about the housekeeping tasks of programming. They might, in fact, be wiz coders but they never use source control and don’t know how to build an executable. Merging forks, installing MacPorts, or editing makefiles will confuse them, and steer their focus away from the interface Gestalt toward the programming equivalent of dishwashing. Programmers are used to doing their own dishes and forget that these are actually complicated tasks that bear little relationship to what you want your interface designers actually doing with their donated time: designing interfaces. Volunteer to help them with these tasks to keep them focused on the things you need them to do.


I know I’m hardly the first person to think this stuff through. John Gruber’s 2004 article Ronco Spray-On Usability is a classic, written in response to Eric Raymond’s The Luxury of Ignorance: An Open-Source Horror Story. David Nichols and Michael Twidale wrote probably the most thorough analysis in 2003, and I borrow many of their themes here. The OpenUsability project in many ways arrived at exactly the opposite conclusions I propose here. Celeste Lynn Paul offers another set of suggestions and provides a little historical context.

The writing on 大黑狗 tends toward the personal and away from the professional. When I write about my job, it’s usually in a “funny thing that happened at work” kind of way. I’m not a pontificatiforous person when it comes to my job. They pay me to push pixels so I push the pixels, and I try to do it with integrity and enjoyment. But as I dig more and more into the FOSS world — I’m running Ubuntu on all my Macs now, and I’ve switched to MacVim fulltime, for example — I’m consistently horrified by the lack of attention paid to interfaces. The FOSS A-listers get this, they know there’s a problem, and they talk about fixing it. But I think they miss that design (graphic, interface, whatever) operates on a totally different paradigm.

I don’t write on 大黑狗 for traffic or fame or money. I literally write it for my friends and family. I would love love love if this post got a little traction in the FOSS community. If you agree, or disagree, or have something to add, or know of similar comments elsewhere ... fire away. But in about two weeks of reading about FOSS usability, I’ve yet to see anyone come out and say it as bald as this: interface design isn’t actually much like programming. The root of the FOSS usability problem is expecting the opposite.