Why HTTP/3 Matters: Speed Gains, Benchmarks, and Real‑World Adoption

HTTP/3 is now officially standardized as RFC 9114, announced on June 7 2022 after five years of development, marking the third major version of the Hypertext Transfer Protocol.

At the same time, HTTP/2 was updated to RFC 9113, and the new HTTP/3 standard works together with RFC 9204 (QPACK header compression) and RFC 9218 (extensible priorities). The cURL project has already made HTTP/3 the default.

Benchmark Results

A benchmark was run using the same browser, network, and site, varying only the HTTP protocol. Each site was fetched 20 times and response times were measured via the Performance API.

The chart below shows clear performance improvements with each newer HTTP version, especially over longer distances and less reliable networks.

Brief History of HTTP

HTTP 1.0 was completed in 1996, followed by HTTP 1.1 in 1997 to address issues such as proxying, caching, persistent connections, and virtual hosting. After a long gap, HTTP 2 was standardized in 2015 (RFC 7540).

“HTTP/1.0 did not adequately consider layered proxies, caching, persistent connections, or virtual hosts, and many incomplete implementations called themselves ‘HTTP/1.0’, prompting the need for a new version.”

One File at a Time

In HTTP 1.1 a browser could download only one file per TCP connection, causing head‑of‑line blocking. Opening multiple TCP connections mitigates this but consumes more resources, especially when TLS is involved.

Multiplexing in HTTP/2

HTTP 2 introduced binary framing and multiplexing, allowing a single TCP connection to download multiple resources in parallel. However, TCP‑level head‑of‑line blocking remains, and packet loss can stall all streams.

“Because TCP’s loss recovery is invisible to HTTP/2’s multiplexing, lost or reordered packets pause all active transactions, regardless of which stream they belong to.”

In high‑loss environments HTTP 1.1 can sometimes outperform HTTP 2.

True Multiplexing with HTTP/3 and QUIC

HTTP 3 replaces TCP with QUIC, a UDP‑based transport that provides stream multiplexing, per‑stream flow control, and integrated congestion control, eliminating TCP‑level head‑of‑line blocking.

Why HTTP/3 Is Faster

True Multiplexing

Because QUIC’s streams are independent, no head‑of‑line blocking occurs at any layer, which is especially beneficial over long distances and lossy networks.

0‑RTT: A Game‑Changer

HTTP/3 supports 0‑RTT QUIC connections, reducing the round‑trips needed to establish a secure TLS session.

“0‑RTT allows a client to send application data before the handshake completes by reusing parameters from a previous connection.”

However, 0‑RTT sacrifices forward secrecy and is vulnerable to replay attacks.

Security considerations for 0‑RTT: Data is not forward‑secret because it is encrypted only with a key derived from the provided PSK. Replay between connections is not prevented.

Can You Use HTTP/3 Today?

The protocol is standardized and already implemented by many services. NGINX offers experimental support, Google and Facebook serve traffic over HTTP/3, and Windows Server 2022 includes HTTP/3 support with additional configuration.

Conclusion

HTTP/3 can significantly improve user experience, especially for sites that load many resources. As the standard matures, it is advisable to start planning its adoption.

Edited by Luo Yi. Source: https://requestmetrics.com/web-performance/http3-is-fast