From the "While You Were Out" Department: I'm one of the few tech-oriented bloggers who didn't note the BumpTop user interface prototype from University of Toronto shown at TED a few months back. That's what comes of being nose down in a startup with a different focus, but it's ironic given that I was an author on related work done during my Apple days and on the related patent (fair warning there). Mind you, while I had design and directional input, most of my contribution was of the "provide resources and stay mostly out of the way" managerial sort.
The BumpTop research group's site (scroll down) gives pointers to a CHI paper and thesis with further background on the design and related prior art related to organizing information in a 'non-desktop' fashion. The original desktop design, born at Xerox PARC, was designed to manage a few hundred items, according to Alan Kay and others who were there at inception. Ordinary users exceeded that threshold long ago, and the situation has gotten far worse since we struggled with it in Apple research in the early 90's.
The natural instinct of the computing industry is to tackle intractable problems by throwing cycles at them, since we can always count on having more next year. Search-like features such as Spotlight on OS X are attempts to leverage compute power to deal with complexity overrun, but without overturning the underlying desktop metaphor. Bumptop is a big step in a different direction, taking advantage of the lush graphics capabilities of modern machines to step outside the designs canonized back in the days of the Alto and 68000-based Macs, in favor of a more reactive, free-form interaction between user and computer. (You did watch the video, right?) Those of us who suffered through working out the cycle budgets for grayscale bitblts can only drool in envy watching natural color, malleable 3-D representations of documents backed by a game platform physics engine.
Seeing this new mutation of the piles notion recalls another idea from back in the day. Randy Smith, now at Sun, wrote a paper regarding "the tension between literalism and magic". Though the project he was describing is long obsolete, the analytical framework he introduced is not. Literalism are those elements in an interface design that transfer more-or-less directly from experiences in the real world and therefore promote easy learning. (These concepts translate into marketese as 'metaphor' and 'intuitive'.) Magic is those properties of the interface that could not happen in the real world, and use the underlying computing and communications systems to extend the power of the interface beyond the real world analogues to which allusion is made. The synthetic world can have 'supernatural' powers, but the user will have to learn to control them.
The original desktop design leaned in the direction of literalism. While the allusion to reality was never pure (trashcans on desktops?) the generally one-to-one correspondence between user action and resulting change inside the system put it squarely into the direct manipulation class of literalist designs. This literalism is also a large cause of the failure of the desktop design to scale, as the user is responsible for acting to create and maintain useful organization of ever-growing collections of information. Contrast the wildly successful - and almost completely 'magical' - interface of Google. There is no real world act equivalent to typing a few words and receiving in return lists of information from any place in the global Web. There is an equivalent in the realm of magic, however: "Naming calls." The overwhelming acceptance by end users of an interface devoid of literalist elements is a quiet and widely overlooked revolution of the last decade, and its implications are largely unexplored.
The literalism and magic pattern plays out well against the piles design, all the way back to its origins. The underlying project was a collaboration within Apple Advanced Technology between the Human Interface Group (HIG) then managed by Joy Mountford, and an Information Access Group initially under my own management. One was representation, one implementation - yin and yang as it were - though there was a lot of cross pollination. The same when time came to write up the project. The peer group of HIG resided within the ACM SIGCHI community, and that's where the paper focused on the literal elements of piles was published. Similar the algorithms team published their parallel implementation paper within the SIGIR community, describing the basis of the 'magical' elements of the design.
Which of these threads has been more influential within its community? Easily answered: the SIGCHI paper has 49 cites in the ACM database, while the SIGIR publication shows seven. The Bumptop design falls squarely within the SIGCHI, and therefore literalist, thread - the team's short paper doesn't even cite the SIGIR version of the Apple piles work. And indeed even a cursory look at the Bumptop design shows this is so. The emphasis is on visual emulation including real world physics, integration of UI features from the last decade, while there are fewer algorithmic - magical - properties than our 15 year old Apple work. So while the Toronto team is taking full advantage of the progress of Moore's Law to attack the literalist shortcomings of our long-ago efforts, it has if anything backed away from applying interim progress in information organization to create magic in the interface. By doing so, it's also failed to make a large dent on the scaling limits that afflict literalist designs in a burgeoning, networked information space. That opportunity remains open.
(Thanks to Will Fitzgerald for catching a typo in the YouTube link.)