Rumour goes that a professor once said to his class that they'd get an A+ in a course if they could invent a better compression method than dictionary compression(?), believing the problem to be impossible.

Mr Huffman was sitting in the audience, and actually achieved it, creating 'Huffman Coding' and an A+ :)

HuffmanCoding is a PrefixFreeCode?, which means that no code is a prefix of another one.

an example HuffmanCoding might be |0| A |101| B |11|C

AAAAAACAACAABA

would be encoded as: 00000011 00110010 10

for a total of 18 bits, instead of 28 bits if you used 2 bits per charactor.

ArithmeticEncoding was invented later and allowed you use partial bits to store charactors, but has been under patent by IBM.