[182] discusses aspects of data set size, computational
feasibility, and in particular limits of visualization for ''large''
(about 
bytes) and ''huge'' (about 
bytes) data sets,
where ''large'' and ''huge'' are terms introduced in Huber's
taxonomy of large data sets (Huber, [93],
[94]). As pointed out in [182], even in the most
wildly optimistic scenario, i.e., an angular resolution of 4.38
minutes of arc as suggested in [112], immersion, and a 4:5
aspect-ratio, the human eye would only be able to distinguish
pixels. Using
single-pixel coding, it seems to be impossible to visualize
''large'' to ''huge'' data sets.
[94] initially suggests to prepare ''medium'' (about
bytes) derived data sets that are easier to visualize and grasp
as a whole and can still be worked with established techniques of high
interaction graphics. Unfortunately, as [182] further describes,
common ways of parsing data sets down, e.g., clustering, discriminant
analysis, and principal components, are computationally complex (often
of a magnitude of
or even 
) and therefore are not
valid alternatives. It seems that simple random thinning is the only
methodology of choice, but this may have the side effect of missing
some of the tail structure an analyst may actually be looking for.
Possible solutions are the use of 3D VR
techniques that may display up to voxels ([182]),
further advances in selection sequences, or new strategies to increase
visual scalability i.e., the capability of visualization tools to
effectively display large data sets ([71]). However, the
conclusion in [182] that ''visualization of data sets
say of size or more is clearly a wide open field'' is still
valid today.
Historically, one of the problems with interactive and dynamic statistical graphics was to publish visible results. The ASA Statistical Graphics Section Video Lending Library was one attempt to capture at least some snapshots of software and applications of interactive and dynamic statistical graphics and preserve them for the future. Publishing a sequence of screenshots in a written paper clearly has not the same effect as watching the full interaction and being able to manipulate the graphics.
Recently, many books have been published with accompanying CDs and DVDs and many conference proceedings have been published on CD. Due to this trend, it is now possible to immediately publish a movie accompanying a written paper or integrate interactive graphics within a paper. Examples are [199] and [160] where the papers are accompanied by several movie segments. It is expected that more and more future publications on interactive and dynamic graphics will be accompanied by an interactive application or by movies.
To be useful for the future, interactive and dynamic graphics have to adapt to challenges posed by ''large'' and ''huge'' (in terms of Huber's taxonomy) data sets as outlined in Sect. 10.7.1. Examples for such data sets are data from earth or planetary observation systems, real-time stream data (such as from surveillance systems), or any other type of massive data sets. [183] provides a futuristic vision, indicating that major advancements can be expected in DM, visualization, and quantization methods.
For smaller data sets (from ''tiny'' to ''medium'' in terms of Huber's taxonomy) we can expect to see progress in new user paradigms that will allow to interact with the data through voice or gestures as well as multiple users to manipulate the visible view simultaneously. It can also be expected that more graphical software will make use of the Web and mobile data transmission and reception techniques. The same software may therefore be available for a variety of hardware platforms with screens as small as a clock or cell phone or as big as a 3D IMAX theatre.
As pointed out in [30], ''there are many barriers to acceptance of new methodology by federal agencies.'' This can be easily extended towards other users of newly developed interactive and dynamic graphical software. Clearly, just promoting a new idea or graphical software product is not enough in many cases. It is likely that more usability tests of new graphical software products as well as comparative reviews of old and new tools will be conducted in the future.