In the early days of wireless, spark transmitters were all they had. Selectivity and bandwidths were not major concerns. These days it is all so different with sometimes very crowded bands and the bandwidth of both transmitter and receiver being of major importance.
See http://en.wikipedia.org/wiki/Spark-gap_transmitter for some interesting reading. I recommend you do not try a spark transmitter as you would be very unpopular because of the wide signal and interference caused. By all means read about them though.
A modern challenge is to see just how simple rigs can be made yet still function credibly on the amateur bands today. Rigs like the Pixie are fun, but such ultra-simple transceivers are let down by deaf or easily overloaded receivers. To my mind, the receivers have to be selective, sensitive, have netting, and not be easily overloaded. The challenge is to overcome these issues! Often TX power is not the deal breaker.
UPDATE 1740z: W5OLF has just shown me a photo of is tiny 1 inch ferrite rod antenna on which he has had some success with on 10m WSPR. I tried some WSPR experiments with ferrite antennas some years ago on 40m, 30m and 20m. As long as the ferrite does not saturate they do work. See www.g3xbm.co.uk .