HTTP/3 and QUIC explained: what actually changed for your site
HTTP/3 maps HTTP semantics onto QUIC over UDP. Independent streams kill head-of-line blocking, connections migrate across networks - but UDP can be blocked.
HTTP/3 is the third major version of the protocol the web runs on, and the first one not built on TCP. RFC 9114, published in June 2022, describes it as a mapping of HTTP semantics over QUIC - a transport protocol that runs over UDP instead. If your site sits behind a major CDN you may already be serving HTTP/3 without having decided anything; this guide explains what actually changed and what is worth checking in your own stack.
It follows RFC 9114 and platform documentation, checked on 19 November 2026.
Why QUIC exists: the head-of-line problem
HTTP/2 multiplexes many requests over one TCP connection, which fixed a lot of HTTP/1.1's clunkiness but kept one flaw: TCP delivers bytes strictly in order. One lost packet stalls every stream on the connection until the retransmission arrives - head-of-line blocking, now at the transport layer where the browser cannot work around it. QUIC moves streams into the transport itself: each stream is delivered independently, so a lost packet stalls only the stream it belonged to. RFC 9114 also describes QPACK, a variation of HTTP/2's HPACK header compression that gives the encoder control over how much head-of-line blocking compression itself can cause.
What else changes in practice
- Faster connection setup. QUIC folds the transport and TLS handshakes together, so encrypted connections start with fewer round trips - most noticeable on high-latency mobile networks.
- Connection migration. QUIC connections are identified by a connection ID, not by IP address and port, so a phone moving from Wi-Fi to cellular can continue the connection instead of restarting every transfer.
- Encryption is not optional. QUIC integrates TLS; there is no unencrypted HTTP/3.
The honest caveats
- UDP is not universally welcome. RFC 9114 itself notes that blocking UDP can prevent establishing a QUIC connection, and says clients should fall back to TCP-based HTTP versions in that case. Some corporate networks and middleboxes filter or throttle UDP; fallback behavior is part of the design, not an edge case you must engineer yourself.
- It is not automatically faster for you. HTTP/3 helps most on lossy, high-latency links. On fast stable networks the difference is small. Measure with your real audience before crediting it with improvements.
- Your CDN decides more than your origin. Most sites get HTTP/3 by enabling it at the CDN edge. The origin connection often remains HTTP/1.1 or HTTP/2, which is fine - the user-facing hop is where the transport gains apply.
What to check on your own site
Confirm whether your CDN or host offers HTTP/3 and whether it is enabled, then verify from the browser side: developer tools show the protocol per request. If you serve your own edge, confirm UDP port 443 is reachable from your users' networks and that your fallback to HTTP/2 works when it is not. For the rest of the protocol toolbox, see HTTP redirects explained and HTTP caching for business sites.
How HTTP/3 is discovered
Clients typically learn that a server speaks HTTP/3 through the Alt-Svc response header or HTTPS DNS records after an initial TCP connection - the first visit usually starts on HTTP/2 and upgrades on the next. That means a fresh test from one browser session can under-report HTTP/3 usage, and disabling it is rarely instant either. Keep this in mind when measuring before-and-after performance.
Sources
- RFC 9114, HTTP/3 (June 2022): https://www.rfc-editor.org/rfc/rfc9114.html
Drafted with AI assistance from this source. HTTP/3 support varies by CDN, hosting platform and client network; verify behavior for your own stack and audience rather than assuming availability.