If p(i) (i = 1, ..., N) is the probability of the ith letter of a memo
ryless source, the length l(i) of the corresponding binary Huffman cod
eword can be very different from the value -log p(i). For a typical le
tter, however, l(i) almost-equal-to -log p(i). More precisely, P(m)- =
[GRAPHICS] , where c almost-equal-to 2.27.