For the information geek, we need a generic filter that sends the text fragment to any pre-defined system command, and sets the rendered HTML output as the output of the command. Graphviz Filter is a special case of such a filter, but we really need a generic system.
I'm thinking: gnuplot, LilyPond, LaTeX, etc. All housed under one roof.
Some design issues arise: Each of these programs produces output in a variety of formats.
The taxonomy system is one of the great contributions of Drupal to the technology of CMS: the ability to classify content with multiple vocabularies. Each vocabulary represents a dimension of concern in the content being managed. Further, a single taxonomy has a flexible structure: it can be linear, tree-like, even graph-like when multiple parents are chosen. Content can be restricted to be tagged with a single term, or multiple terms can be allowed. Finally, freetagging is supported.