That makes sense. I think fundamentally you’re trading off space between the compressed data and the lookup tables stored in your decompression code. I can see that amortizing well if the compressed payloads are large or if there are a lot of payloads with the same distribution of sequences though.