From: Alan Wexelblat Subject: Re: APPS: Using VR to represent complex hierarchies Organization: Advanced Human Interface Group Keywords: help wanted This thread reflects an ongoing confusion in the VR community, largely due, I think, to an ignorance of user-interface research. Forgive me if I pontificate a bit here... The problem has to be divided into two parts: representation and visualization (or presentation, more generally). The work cited (PARC cam/cone trees) mixes these two. Fisheye views (Furnas' 1986 CHI paper on the subject is still the best) are pure presentation. A bit of research has been done in this area -- probably the best known is the PARC Model-View-Controller (MVC) architecture. This has been written up in a bunch of papers. At MCC a few years back Kim Fairchild, Greg Meredith and I put together a more advanced version of MVC which we called Automatic Icons (published in Interacting With Computers V1#2). Automatic icons reified the MVC model and gave a mathematical basis for linking prepresentation and presentation. An underlying theory for auto-icons was published in my Semantic Spaces chapter in Benedikt's CYBERSPACE book. This chapter talks about pure representational spaces. Related issues are also discussed in Fairchild's chapter in my recent book VIRTUAL REALITY: APPLICATIONS AND EXPLORATIONS. Recently, the ideas of auto-icons have been independently re-invented by Ralph Hill at Bellcore under the name Abstraction-Link-View (ALV) as presented at CHI'92 and several recent pubs. Ralph and I have talked about this and I think his ALV is fully as powerful as auto-icons, and he has a stronger software substrate supporting his work than we did. Now, with that background, let's turn to the problem: VR, loosely construed, can be seen as a set of presentation technologies. NOTHING in VR is going to solve problems of representation. At best, VR technologies will allow you to make available presentations that would have been awkward or impossible without the technologies. But if you do not have a good, solid, generalizable, scalable(*) representation underlying your presentation, no set of technological tricks is going to help you. (*)This, in my mind, is the beauty of the original Furnas paper on fish-eye views. He shows that the technology can scale almost indefinitely and certainly over several orders of magnitude. By contrast, the cone/cam trees and perspective wall presented by PARC do not scale and -- like most presentations -- tend to break down around 1000 nodes.