Understanding HTTP/2: Its Achievements, Limitations, and the Need for HTTP/3

In recent months many major companies have begun deploying HTTP/3, making it clear that the protocol is on the verge of widespread adoption. As front‑line developers, it is time to understand what HTTP/3 is and why it is needed.

This piece first revisits HTTP/2, which was standardized in May 2015 (RFC 7540) and saw broad browser support by the end of that year. Although adoption grew around 2018, the IETF officially approved HTTP‑over‑QUIC as HTTP/3 in November 2018.

HTTP/2 Highlights

HTTP/2 introduced multiplexing, binary framing, header compression, and server push to improve efficiency over HTTP/1.1. However, it still relies on TCP, which brings its own head‑of‑line blocking problems.

While HTTP/2 eliminates HTTP‑level head‑of‑line blocking by allowing independent frames to be sent out of order, TCP still delivers packets in sequence. If a single packet is delayed, the entire TCP stream stalls, causing all multiplexed requests to suffer.

TCP‑Related Issues

TCP’s three‑way handshake consumes 1.5 RTT, and HTTPS adds another RTT for TLS, resulting in 3–4 RTT before data can be exchanged. On high‑latency links this can add up to a second of perceived delay.

Upgrading TCP itself is difficult because countless intermediate devices (routers, switches, firewalls) and operating system kernels would need to support the changes, leading to the concept of “protocol ossification.”

Why Not Upgrade TCP?

Replacing TCP would require coordinated updates across all network infrastructure, which is costly and slow. Consequently, many TCP enhancements remain undeployed.

Abandoning TCP?

Given TCP’s limitations, the only practical path forward is to adopt a different transport. UDP, being connection‑less and not subject to the same ossification, serves as the foundation for QUIC, the transport layer used by HTTP/3.

Conclusion

HTTP/2 solved HTTP‑level head‑of‑line blocking but cannot overcome TCP’s inherent latency and ordering constraints. Because upgrading TCP is impractical, HTTP/3 leverages QUIC over UDP to provide faster, more reliable connections, setting the stage for the next generation of web performance.