Off the Beat: Bruce Byfield's Blog

Icons have always intimidated me. Except for the mouseover help, two-thirds of the time I would have no idea what function they represent. Shrink them so that they fit on a toolbar, and the obscurity is compounded by illegibility. On the free and open source software (FOSS) desktop, icons seem to be one of the last holdouts against usability, with neither of the two main strategies for designing icons being particularly successful.

Admittedly, icons on the FOSS desktop have come a long way since the early years of this century, when GNOME's logout button looked like an illustration from Goodnight Moon, or possibly a sign indicating an outhouse.

That icon disappeared when the first usability study showed that people had no idea what it indicated, to be replaced (more sensibly) in many distributions with an open door and an arrow or a person passing through it to suggest to leaving. Similarly, when I open up the copy of OpenOffice.org that comes with Debian, I see icons for font weights, alignment and indentation whose functions are obvious at a glance. Other icons, such as the letters ABC above a check mark for spellchecking, or binoculars for search, are less immediately identifiable, but still more or less decipherable.

However, such improvements seem to be only the most obvious and most easily iconified features in any applications. Open any application, and you are likely to stumble across icons whose connections to the features they invoke is either obscure, or recognizable only after a moments' thought. For instance, what connection do the developers of VirtualBox imagine between a cog and Settings? Why does the Krita team use three overlapping circles for Paint with Filters or the maintainers of KDE's System Settings between what looks like a pad of paper with a diagonal slash through it for Appearance?

It gets worse, too. I have yet to decipher what the icon for Browse represents in Okular, or the one for Properties in digiKam.

And these are not necessarily the worst offenders -- just random examples taken from the applications I run regularly. Nor are these examples meant to condemn a particular desktop or application, although I have noticed that newer GNOME applications tend to have more easily identifiable icons than most.

Regardless of the desktops or applications, I suspect that a lot of people are like me, and only work with some of these icons by remembering their position on the toolbar -- especially in applications like The GIMP, with its thirty sub-microscopic icons. Clearly, there is not only rooms, but entire country estates left for improvement.

Abstract reductions

One way that designers have tried to address the usability problem is by simplifying the icons to a few basic shapes. This tactic has the advantage of producing clear images even at small sizes yet by itself is not enough to produce usable icons.

For instance, in the latest Ubuntu, the Internet applet on the panel is two arrows, one pointing up and one pointing down. Command line users will recognize the design as a reference to the ifup and ifdown commands used to configure ethernet connections, but, to the desktop users that Ubuntu targets, the reference will be meaningless.

The same is true of the orange starburst update icon in the notification of Fedora and several other distributions. The icon suggests a matter of urgency (which usually it is not; at least, not a matter that must be attended to this very second), but what is so urgent is not conveyed at all. The same lack of identifiable purpose is visible in the rectangle and a plus sign to mean Open in several newer applications such as GNOME's Simple Scan. Simplification is a worthwhile goal in a media as limited as icons, but if the design is over-simplified, then all the designer has found is a new way to be obscure.

Nostalgic references

The second major tactic in icon design is to refer back to older technologies. For instance, the floppy drive has been dying slowly for over a decade, and has been not used as a main store device for twenty years. So why is it still being used in many applications for the save icon?

I see that the latest version of Ubuntu has tried to update this icon by showing an external drive, but it is harder to depict clearly in a few dozen pixels than a floppy drive, and the reference is weakened by the fact that an external drive is what most people use for back ups, not for saving files.

In the same way, the Lancelot menu in KDE (to use another random example) uses three obsolete cultural references for its top level menus. Computer is indicated appropriately enough by a laptop, but Applications are indicated by a CD leaning against a box -- never mind that this is free software, and applications rarely come in boxes. Lancelot also uses a stack of letters for Contacts, and a typewriter for documents. Some of these same icons can be found in other KDE applications as well.

Such cultural references may be tempting to designers, because they seem a quick way to solve the problem of how to produce a recognizable icon. But the trouble with such references is that, as time passes, they become less and less meaningful. I am old enough to remember (and bless!) the transition from typewriters to word processors, but today I do not immediately associate typewriters with documents.

Nor does snail mail come to my mind when I think of an address book. These days when I use the word "address," I usually mean an URL or email address. When I want to refer to a postal address, I usually qualify it as a street address.

If such icons hold such little meaning for me, how much less can they mean to a twenty something who has never used a typewriter, and possibly never seen one in person? For young adults, even a recently obsolete technology like floppy drives may not carry much meaning. And when, as in the GNOME desktop of the latest Ubuntu, an envelope represents not only email, but also messaging, the analog is so remote to many users that it might as well not be used.

What seems at first to be a convenient shortcut on second glance becomes only slightly less arbitrary than a randomly selected symbol.

Rethinking icons

Compared to the task of improving functionality, developing meaningful icons might seem trivial. Part of me must agree, since, despite all the complaints I make here, I still manage to accomplish my necessary tasks. All the same, if usability needs to be a priority if the FOSS desktop is to become widely used, then icons are one of the obvious starting points.

Right now, I can find very few applications which have a complete set of default icons that are usable. Firefox does. Brassero very nearly does, except that the common element of a disc is far larger than the unique element that defines each function. However, most applications have a mixture of clear, distinct icons and ones that fail for one reason or the other.

In fact, modern applications, proprietary and FOSS alike, seem to me explicit admissions that icons do not work very well. Many applications provide mouseover help to help users identify the icons. The KDE desktop goes one step further, offering thumbnails to help you see what each icon does or contains. If icons worked the way they are supposed to, then why would these supporting structures even exist?

Many icons are simply inherited, like a man page that passes from operating system to operating system without being updated. Yet it seems to me that the FOSS community either needs to get even more serious about icon design that it currently is, replacing the icons that fail to convey information, or else consider if icons are worth having at all.

True, icons contribute strongly to the look of a desktop or an application, and doing without them would seem strange. But they are redundant, since the same functions are available from the menu or via keyboard shortcuts. So, if most of them are not functional, why keep them simply for the sake of tradition? Icons may have been originally designed to help users navigate, but, if so many of them are unreliable markers, then either they need to be seriously improved, or perhaps eliminated altogether.

Comments