The Request Tax: Re-evaluating 20+ Years of Web Performance Dogma
summary
Benchmark data shows HTTP/3 and QUIC cut per-request overhead so that many small fetches beat one large bundle: measured curves put HTTP/1 performance rising (worse) as payloads split while HTTP/3 stays flat and faster, especially under moderate latency; charts use red boxes for HTTP/1 wins on very large single payloads and green for HTTP/3 wins as chunk count increases. Recommends HTTP/3 and QUIC reduce per-request overhead and to prefer caching over aggressive bundling and calls for real-world Cloudflare/Astro data.
AI-written summary · about
chapters (9)
- 0:00An Intervention for Web Developers
- 1:26Why We Started Bundling
- 2:24Bundling vs. Caching
- 3:07HTTP/2, HTTP/3, and QUIC
- 4:09What the Performance Data Shows
- 5:40Stop Being Afraid of Network Requests
- 6:09Optimize for Caching First
- 6:55It's Time to Rethink Bundling
- 7:45What We Need from the Industry Next
related talks
51mHTTP/3 and QUIC: Who, what, where, when and, WHY?
30mHTTP/3 Is Next Generation HTTP. Is It QUIC Enough?
46mHTTP/3: It's all about the transport!
34mQUIC and HTTP/3: Too Big to Fail?!
37mHTTP/3: why and where are we?
20mHTTP/3 in Envoy: End to End Acceleration
40mHTTP/3 - HTTP over QUIC is the next generation
35mOn QUIC, HTTP/3 and tech stack unification