It's always seemed surprising to me that variable-length codes (Huffman, Shannon, etc.) are conceptually more complex than replacing repeated sequences (LZ family), yet the former were "discovered" first. Even RLE, which is like a very limited subset of LZ, came after Huffman. If you compare the amount of code needed to decode even simple static Huffman vs. a common LZ variant like LZ12/4, as well as the typical compression ratios of those schemes, it's clear that the latter is a huge win in efficiency and simplicity. Maybe the lesson here is that if you take something that seems almost like common sense or trivial and turn it into an invention, you just might get lucky and become famous?