My feelings regarding form over content and the Tragedy of HTML are in agreement with Abigail's long-standing manifesto cited above. The CERN/TB-L view of the WWW was ruined by the introduction of Marketing-style graphic design. However, when the data is rich enough, graphics and even color graphics are appropriate. See Bertin and/or Tufte. 1 2 3 4
(Note that conformant to dogma, these are em tags not i tags!) Abigail's manifesto is upheld perfectly, since not only was the data rich enough to grpah, the article showed the code but didn't link the graphic.

system()ing out to gnuplot is fine for those who have it and are comfortable with it. I recall doing stereograms with Unix V7 plot and a daiseywheel terminal, it works, but I've not played much with gnuplot. The example is even readably self-documenting as an embedded here-doc, I like that. (I may be over-sensitized on that; Uri was sounding off on readability last night in his Boston.PM rehearsal of his YAPC talk.) Hiding the system() in a open("|gnuplot") that demos 5.6's autoviv FDs is even cooler.

In spirit of TMTOWTDI and CUFP, I rather like Stein's GD.pm and Verbruggen's GD::Graph wrapper for Boutell's libgd. Latest revs support TrueColor as well as palettized, and both PNG and JPeG (GIF is gone). I see there's even now a Template::Plugin::GD::Graph in the Template-Toolkit, which could be interesting for on-the-fly graphics (which I try to avoid, but could be useful if the data xor users are dynamic).

Cheers,
-- Bill N1VUX
If I post a few more times I should register, eh?