A deep technical conversation with Fabian Giesen covering the history and internals of lossless data compression. Topics include the LZ77 and LZ78 algorithm families, why ZIP's 32KB window limit exists (DOS memory constraints), how Zstandard improves on ZIP via parallel bitstreams and wider copy operations, and how Oodle (UODL) further optimizes for game data with 3–6 parallel streams and structured binary data predictors. A significant portion covers software patent absurdity in compression: how PKzip, ARJ, and other archivers patented trivial data structures (hash tables, linked lists, binary trees) rather than actual compression innovations, how the gzip/InfoZIP team deliberately chose a patent-free linked-list structure and published it as prior art, and how codec patent pools like H.265/HEVC lead to ongoing litigation (Nokia vs Acer/ASUS). Compression metrics discussed include ratio, encode speed, and decode speed, with the key insight that LZ77-class algorithms are asymmetric — decompression is far faster than compression.
Sort: