Furnas, G. W., & Bederson, B. B. (1995). Space-Scale Diagrams:
Understanding Multiscale Interfaces. In Proceedings of Human
Factors in Computing Systems (CHI 95) ACM
Press, pp. 234-241.
By Lance Good
Much of this paper should seem familiar. Many of the figures and
explanations appear in the Pad++ paper we read in the previous
class. You can view this paper as a "zoomed" view of the
Space-Scale Diagram section of the Pad++ paper (ha-ha).
The motivation given for Space-Scale Diagrams is to better
understand Multiscale 2D Interfaces. This includes almost all
the interfaces mentioned in class thus far including Pad, Pad++,
Fisheye views, etc. The paper begins with an explanation of Space-Scale
Diagram concepts followed by some sample applications of these
diagrams to Multiscale Interfaces.
Space-Scale Diagrams
Since we have already read about these diagrams, I will only
cover some of the possibly less intuitive concepts.
An intuitive explanation for the shear property is to consider the
given surface as only part of an infinite plane. If we considered
the surface, say, as only the upper right quandrant of a surface
four times the size, we get the diagram shown in figure 4.
A way to think about the (x,z) coordinate system is that x gives
you the center of the zooming box and z determines the size of the
zooming box. This zooming box is then stretched or squeezed to
the size of the screen. The (u,v) coordinate system, on the other
hand, fits with the Space-Scale Diagram. The u gives you the
position of the viewing window and v gives you the height.
Applications
With the joint pan-zoom problem the diagram in Figure 7 helps you
see that the diagonal of the parallelogram might be a good solution.
In case there was any confusion about the formulas:
m = (v1 - v2)/(u1 - u2)
(slope = rise over run)
m = (z1 - z2)/(x1z1 - x2z2)
(change coordinates)
(v - v1) = m(u - u1)
(basic formula for a line)
z - z1 = m(xz - x1z1)
(change coordinates)
z = (z1 - mz1x1)/(1 - mx)
(solve for z)
The basic thing to come away with from the sections on pan-zooms
and shortest paths in scale-space is that finding the optimal
solution is difficult. For instance, the above formula can
always be used, but when is it optimal? As the parallelogram of
Figure 7 approaches a horizontal line, the situation becomes much
like Figure 8, a pure pan. There are many cases to consider here
and as the paper mentions, finding optimal solutions is complicated.
Further, it is not clear whether the best solution on paper is the
best empirical/cognitive solution.
The next idea, semantic zooming, seems to be essential to Pad++
and JPad. Without semantic zooming, a hierarchy structure could
not be represented. This section also introduces a fractal grid
that works on this concept.
The paper concludes with a description of various distortion
techniques to which Space-Scale Diagrams possibly bring further
understanding. Certainly, Space-Scale Diagrams make such distortions
easily describable.
Things I Liked
The diagrams were very helpful. I especially appreciated
having examples of both 3D and 2D diagrams.
The fractal grid and distortion views were nice examples.
Issues and Things I Might Change
I understand the arguments for monotonicity and the idea of the
"bounding parallelogram". I thought a clear explanation or generality
of when monotonicity is good and when it is bad would have helped. For
instance, the next section contradicts monotonicity (see Figures 8 and 10)
and it never seems to surface again.
I guess I could read the paper, but I was wondering how the
Fish-eye view was explained in 1982, eg. without Space-Scale Diagrams?