Honey, I Shrunk the Bits. Or how modern HTTP compression works.
summary
Gzip stays ubiquitous via zlib and deflate with typical text gains around 70–90% and uses the standard Content-Encoding/Accept-Encoding headers, Brotli wins ~10–20% smaller payloads for web text by using a 120KB static dictionary and the BR token, Zstandard (Facebook 2016) targets real‑time streams with Zstd: faster than gzip, near Brotli ratios using finite‑state entropy and table lookups, and practical choices hinge on build vs runtime compression, CDN caching, and whether a custom dictionary can be shared.
AI-written summary · about
chapters (25)
- 0:00Introduction
- 1:07Why compression matters
- 2:34Sci-fi inspiration
- 5:49Run Length Encoding basics
- 7:14Huffman encoding
- 9:19LZ77 and LZ78 algorithms
- 11:06Deflate algorithm origins
- 12:30Dictionary-based compression
- 13:32Lossy vs lossless trade-offs
- 14:59HTTP compression headers
- 16:24GZIP format adoption
- 17:51GZIP in HTTP requests
- 18:58Why files benefit
- 19:20Brotli compression
- 21:51Brotli levels and HTTP/2
- 22:33Brotli static dictionary
- 23:16Zstandard compression
- 25:00Zstandard’s Finite State Entropy
- 26:48GZIP vs Brotli vs Zstandard
- 27:50Future compression algorithms
- 28:11LZ and Zstandard deep dive
- 30:40Streaming compression
- 31:00Branch prediction speed
- 33:10Compression impact on performance
- 34:13Q&A
related talks
47mWeb Compression: How to (Re)Take Control?
46mNot Only Gzip
03mWhen small is better: Brotli compression in a real case
52mCompression Dictionaries
15mBeyond Gzipping
51mWeb Compression: How to (Re)Take Control?
24mHow GZIP compression works
36mBreaking The Bloat: Achieving The Fastest Navigations on The Web