Usability

Familiarity

The holy grail of usability is for the system to act, for each user, just like they would expect it to work. So people remarking on the "usability" of Gnome may complain about the icon and "Applications" menu being all part of one big menu title, because they're used to icons being distinct from text. Also, the menus have no outline to make explicit that both logo and text are one item.

But I'm not sure this menu thing is really a major problem. It surprised me once or twice, but not since. It's something you get used to with a little practice, sort of like the interface changes in a new version of Firefox. Is it less "usable" because you're familiar with a slightly different interface? You'd probably score slightly worse on a usability test, even though it has been "improved", until you get used to the new layout.

The major problems are in areas where users don't just get used to it. The Gimp toolboxes in Gnome have constantly gotten in my way, floating above the image, being un-minimizable, and other such nonsense. The other day I minimized one of two image windows and the toolboxes minimized with it. Convenient! I was just about to access them for the other image, actually.

Rant Mode: On

(A few minutes ago I found a condescending comment by someone about how the TAB key toggles visibility of all the toolboxes, and doesn't anyone read documentation? Obviously not, but being curious about this allegedly documented feature, I ran grep -i tab on the complete user manual. The only reference to the tab key's toolbox functionality is in the paragraph on fullscreen mode in the "What's new in Gimp 2.4" page of the Gimp 2.6 help*, and the page seems very clear on the fact that Tab is unique to fullscreen mode. I can't find any mention of this magic Tab key in any of the menus, either.
* I guess that means there's nothing new in 2.6.)

The GTK file selector is also constantly tripping me up. Mine is always in open mode, because it's virtually guaranteed to not have the right directory in the listbox. So I'll click in the directory-selector area, and use the keyboard to whip through the hierarchy to whatever location I actually want, then turn my eyes to Name again. Now, that Name box has the X11 selection, so it's helpfully highlighted. But it is also not the input focus, so it's actually an ever-so-slightly de-emphasized version of the real selection color. It's close enough to the proper color that I never realize it's not really selected, even though it's selected, and I try and go into a directory named for my file choice. Or I realize my mistake, and press Escape to get rid of the find-a-directory box, and the whole save dialog closes.

Without remembering my place, of course, so I have to go running through the hierarchy again, possibly making the same mistake. Typically I stop myself before hitting escape, and wonder which key I should press until the find-a-directory box times out and vanishes, at which point I am finally allowed by the grace of Emperor Gtk to hit Alt+N and put in a filename.

Magnitudes

Ranting aside, those are my examples. The worst usability problems a user will face are the ones that go against the grain of that user, to the extent that practice begets familiarity only slowly. Problems are also noticeably worse if there's no workaround. While the Applications menu may be surprising, it doesn't do anything destructive to click on the icon instead of the text, nor does it relentlessly continue obstructing anything.

I would almost say that the menu example is not even a usability problem, but if the other issues were to be fixed, it would rapidly become one. Like reducing the time of something from 1 second to 0.1 second, it doesn't sound like much of a gain, but that small difference can change your entire perception of the system, and maybe even your whole way of working with it.

What if it did change your way of working with the system? Once we become accustomed to a new habit, it's jarring to see the old interface. Like people who get used to searching the Start menu in Windows Vista, then return to a Windows XP machine: all of a sudden they have to worry about the spatial layout of the menu and vendor names. Still, prior to Vista, the old spatial start menu with folders named after vendors ("Google Chrome") were all we had, and we were fairly proficient at using them. Does it make the Vista search box less usable? Or is the XP way less usable in an absolute sense, and shunning the search box in favor of moving around the full programs menu is always holding us back?

Discussing this almost makes me wonder if the way that Vista's start menu is much smaller than XP's, and opens its folders as a tree in that small area instead of a submenu, is a deliberate reduction in usability of the old path in order to nudge users in the direction of using the search box instead. The search results are confined to the same area, but having been searched for, they're of much higher quality and immediately accessible instead of being in Programs→Random Vendor.

As fashionable as it may be, I'm hesitant to regard Microsoft as completely clueless.

Design Goals

How do we really go about measuring usability, when the speed and ease of task completion varies based on experience and familiarity? Usually, the design is optimized for a large number of "average" users, who are using the software for the first time. This masks the effects of long-term experience and familiarity with the specific interface at hand, and ends up testing how well your product acts just like everyone else's.

This sort of consistency is very useful in making your product easy to pick up, but also tends to hold back innovation. If OtherCorp had been developing the iPhone instead of Apple, would they have backed away from an all-touch interface and instead produced something like the touchscreen Blackberry model? After all, very few people at the time used keyless phones, and the phone wouldn't give any tactile feedback on whether the screen registered a press on an object.

The other danger of following strongly established conventions is that your product may be for a professional market, where users have a chance to become real experts with the software. Should Photoshop's default toolbars and feature availability be oriented toward the average corporate graphic artist, or toward people who heard on the Internet once that you need to be able to work with CMYK images for print? (Which I hear, as publishing goes fully digital, is just a needless pain for the actual printer.)

Like security, it doesn't seem that usability is something that can be sprinkled in later. It needs design, and it needs to be designed for someone. Complexity in Photoshop is not necessarily a condemnation of it, but overwhelming complexity in getting your files off your digital camera card is a problem.

Free Software

How does usability fit into free software? One of three things seems to happen: the standard "scratching an itch" example of where a program is built with no thought to usability beyond making the desired features possible; the standard "users are dumb" mode where all possible features and options are stripped out as aggressively as possible; or, rarely, an interface that doesn't hold back, and offers enough options to suit every hand.

Part of the problem is the desire of certain factions to exit the world of the tinkerer and get into the realm of Real Serious Software. They're actively ignoring the actual userbase in favor of chasing a different market, while still claiming they're perfect for everyone.

Also, working on usability isn't very exciting. Once an itch is scratched—once some software exists to solve the original problem—the rest of it required to make a successful project is just too much boring work. And when one isn't getting paid for giving away their work as Free, it makes little sense to spend any effort on the rest of the project.

Finally, usability doesn't get much respect. Witness the amount of flames generated on the release of Gnome 2.0, when everyone's favorite option was no longer configurable. This readily translates to viewing 'usability experts' as clueless academics living in an ivory tower, actively obstructing your lovingly crafted code. (Something I have regrettably been afflicted with in the past.) It also shows in the tendency to complain about every little thing that a reviewer can find wrong with a piece of software, like the logo in the Gnome menubar.