Interfacing the future

he imperfect computer in 2001: A Space Odyssey had what many of us consider to be the perfect interface: omnipresent, always on, watching, listening, and anticipating our every needs. We want a version of HAL 9000, presumably one that won't lock us out of our homes or kill us in our sleep.

Many think that if we're to get there, the first step is to rethink our reliance on the so-called WIMP (windows, icons, menus, pointer) interface. Invented at Xerox's Palo Alto Research Facility (PARC) in the 1970s, popularized by Apple Computer in the 1980s, and milked by Microsoft in the 1990s, WIMP has reached a point where further building on it yields only diminishing returns.

That's not to say it will go the way of the command line; WIMP charmed the world for good reason. However, some say we have relied on its metaphors too literally and too liberally.

As Alan Cooper, software designer and president of Palo Alto-based Cooper Interaction Design, puts it, it may seem clever to use telephone and keypad icons to represent dial-up services, "but the original makers of the telephone would have been ecstatic if they could have created one that let you call your friends just by pointing to pictures of them."

The emergence of the Internet also has shown WIMP's seams. Although we still balance our checkbooks and polish our resumes with them, computers are now information conduits as well as information processors. Steven Johnson, author of Interface Culture and editor of the New York-based online magazine Feed, says he believes that "representing all that information is going to require a new visual language."

EVOLUTION, NOT REVOLUTION. It's unlikely that a new computing interface will emerge into the marketplace as fully formed as Apple's Macintosh OS did in 1984. In the next five years to 10 years we're more likely to see gradual changes to WIMP instead of a full-blown interface revolution.

This already is apparent in Microsoft's controversial integration of a browser into its Windows operating system. A built-in HTML parser isn't a radical interface change, but it is a slender bridge between a computer's dual roles as information processor and information conduit.

Apple also is tentatively spanning the processor-conduit divide with its Sherlock search facility, due to arrive this fall in Mac OS Version 8.5. According to Peter Lowe, Mac OS product-line manager, one of Sherlock's capabilities will allow the Finder, rather than the browser, to access and return results from Internet search engines.

THE SOUND OF THINGS TO COME. Another evolution of WIMP may come through speech recognition. The obvious benefit to speech recognition is that it frees your hands from having to operate the computer: Imagine yourself elbow-deep in blueberries only to realize that you don't remember the required proportions called for in the pie recipe on your computer.

Michael Cohen, Microsoft senior researcher, thinks there could be a side benefit to speech recognition as well. If it catches on, speech recognition may clean up WIMP interfaces by reducing the number of onscreen icons and menus. A hands-free interface also could help curtail computer-related, repetitive stress injuries.

Unfortunately, although dictation engines have improved dramatically in recent years, sophisticated management of OSes has been harder to achieve. Most experts in interface design predict that, at best, speech recognition will integrate itself slowly and selectively into the WIMP interface. Command-and-control functions will likely be an alternative, not a replacement, for the keyboard and mouse.

More importantly, until computers can understand a wide variety of natural phrases, users must remember a specific set of commands in order to use speech recognition. This boxes them in even more than icons and menus do. Social inhibitions also are likely to limit the use of speech recognition. If you're concerned about being overheard, it's unlikely that you will choose to broadcast your actions.

MAKING WIMP WORSE? The oldest element in nearly all computing interfaces is the part that's used more than all the others: the display. Pushing 50 years of age, the capabilities of CRTs and siblings such as LCDs have improved very little compared with the leaps in processor speed and storage capacity. Most displays still don't approach the resolution of an average laser printout. For a premium, a flat-panel device may save room on your desk, but its business end still works in two dimensions.

It's a pleasant diversion to imagine the possibilities if displays obeyed Moore's Law, but there's no telling when holographic technology, which would be a truly exponential advance, might widely be used.

Because it seems we'll be stuck with 2-D displays for a while, simulations of 3-D spaces on 2-D screens have proliferated since the advent of WIMP. None of these 3-D simulations has caught on, though, perhaps because of the hefty system requirements that such tools have. The latest of such 3-D views, including open standards such as Virtual Reality Modeling Language and proprietary technology such as Microsoft's Chromeffects, also will be resource hogs.

There's a potentially larger problem with 3-D views: usability. It's telling that Microsoft code-named its technology Chrome -- high-tech jargon for the silver-gray toolbars and window embellishments that increasingly encroach on precious screen real estate. Microsoft always has overloaded its GUIs, which ironically has made the limitations of WIMP more obvious to the average user.

It's a trap that awaits designers of 3-D views as well. According to Cooper and Johnson, the 3-D interfaces of Microsoft's Bob and General Magic's Magic Cap failed not because they were static or too graphically crude, but because they reproduced the real world too literally instead of finding new metaphors. Even if powered by unimaginably fast processors, fly-through versions of desktops and shopping malls only will make WIMP worse.

INTO HYPERSPACE. One promising new spatial metaphor is the Hyperbolic Tree (H-Tree), a visualization technique invented at Xerox's PARC. The H-Tree has breakthrough potential for two reasons: It implies a space that actually goes beyond traditional 3-D views, and it efficiently does so on a 2-D display without requiring much horsepower.

According to a paper by its creators, the H-Tree "exploits hyperbolic geometry in which parallel lines diverge away from each other. Thus, hierarchies, which tend to expand exponentially with depth, can be laid out in hyperbolic space in a uniform way so that the distance between parents, children, and siblings is approximately the same."

The intent is to create an interface with both focus and context, in which "the relationship of parts to the whole can be seen and focus can be moved to other parts in a smooth and continuous way."

The screen shots on this page give an idea of how the H-Tree looks, but they can't convey the sensation of moving through the structure. Ultimately, the H-Tree is better when experienced than when viewed or described.

Opinions about the H-Tree's potential are mixed. Cordell Ratzlaff, manager of Apple's Human Interface Group, says he believes the H-Tree is of limited use. Feed's Johnson agrees, stating that H-Trees work well with strictly defined indices, such as the Yahoo directory, but are disorienting for chaotic systems such as the Web.

Irwin Chen, information architect at Feed, is more optimistic than his colleague, declaring H-Trees as a "major step toward the dream of perfect information."

Determined to capitalize on PARC's latest creation in a way it failed to with WIMP, Xerox is licensing H-Tree technology through its Inxight Software subsidiary.

The push is paying off. H-Tree technology -- available as ActiveX and Java controls -- has been licensed to several companies, and adoption is spreading. Microsoft, for instance, employs an H-Tree control in its Site Server Content Analyzer application to map and display the structure of Web sites. H-Tree technology also can be found in Softquad's HotMetal Pro, Manage.com's Frontline Manager, and Virtual Integration Technology's delivery Manager. Inxight also has several Java-based H-Tree demonstrations at www.inxight.com/products/hw.

The H-Tree is still in its infancy, technically. For now, it sits at the conduit end of the processor-conduit continuum. An H-Tree might help you locate something or visualize a complex structure, but it can't do much with what it displays.

Of course, none of the technologies mentioned in this article comes without weaknesses. The technologies attempt to advance interface design from different angles, occasionally solving problems but just as often revealing limitations.

It's telling that, when asked to imagine their ideal interface, many interface design experts humbly decline to talk about their area of research, instead falling back on abstract descriptions of knowledge-sharing devices and intelligent assistants. The others just say "HAL."

Chip Brookshaw (chip_brookshaw@infoworld.com) is a senior editor at the InfoWorld Test Center. His beats include handheld computers and wide area networks.