Miki, your warning is spot on.
I try to stay in the "sweet zone" of each data tool used and ask
myself if a graph really requires all data points in order to be
informative. There are times when an aggregation of values will
suffice. For example, if the window of time is one year, it may be
overkill to plot hourly data. I might bring the hourly level time
series data for a 365-day span down to a daily level of data
aggregation, or perhaps compute boxplots for a weekly or monthly data
level, to give a sense of the distribution. Sometimes presenting less
can yield less noise, more signal, and a clearer message.
Also, as you suggest, creating "bins" of data works as well. Gnuplot
is great, though I tend to favor ggplot2 when I'm already working in
R.
I'm relatively new to javascript, but my preferences are definitely
leaning towards browser-based visualizations. See the okcupid blog for
an example of its flexibility/expressivity:
http://blog.okcupid.com/
I wonder if clojurescript (
https://github.com/richhickey/clojure-contrib/tree/master/clojurescript/
) will eventually be a way to achieve some of the innovation occurring
in javascript currently (html5/canvas, jquery, flot, jstat,
processingjs, etc… ?
Cheers,
-Avram