In my ideal world, code would be typeset in proportional fonts, with mathematically appropriate glyphs chosen from a font with a comprehensive Unicode character set, including a proper minus instead of hyphen-minus.

The usual problems with this are arranging indentations/alignments neatly where you really do want them and avoiding an explosion of funny characters once you've got a large part of the Unicode character set to play with. Tools are pretty good at the first these days, but pity the poor Haskell developers if we don't fix the second. People will be writing logging libraries that use different levels of emoticon to save writing 'debug' and 'error'...