Huffman coding was invented back in the 1950’s, and has a key assumption – that we know the expected frequency of each symbol.
We could get this frequency from a large corpus, such as the BNC, but then the resulting huffman tree might not be ideal for all circumstances.
A technical computing paper is likely to contain more non alphabet symbols than ‘war and peace’.
Adaptive Huffman is usable in this situation, as it reconstructs the tree as data is read in.