HTTP/1.1, HTTP/2 & HTTP/3 Compared

HTTP/1.1 — what senior engineers still need

• Request/response over TCP: methods, status codes, headers, body.
• Persistent connections (keep-alive) — reuse TCP for multiple requests; avoids handshake tax.
• Pipelining — largely failed in practice; browsers opened parallel connections instead (typically 6 per host) — domain sharding was a hack to raise parallelism.
• Head-of-line blocking — one slow response blocks the next on the same connection.

Know idempotency: GET/PUT/DELETE vs POST; matters for retries and 0-RTT.

HTTP/2 — multiplexing on one TCP connection

• Binary framing — streams multiplexed on a single connection.
• HPACK header compression.
• Server push — mostly fell out of favor; know it existed.

The catch: TCP HOL blocking still applies — one lost packet stalls all streams on that connection. HTTP/2 fixed HTTP/1 connection limits, not TCP loss behavior.

HTTP/3 & QUIC

• Runs over UDP with QUIC providing reliability, encryption (TLS 1.3 integrated), and per-stream multiplexing — loss on one stream doesn't block others.
• Connection migration — survives IP address changes (mobile).
• Adoption growing at CDNs and browsers; many origins still HTTP/2 or HTTP/1.1 behind TLS terminators.

Comparison table (memorize this)

	HTTP/1.1	HTTP/2	HTTP/3
Transport	TCP	TCP	QUIC (UDP)
Multiplexing	Parallel connections	Streams on one TCP	Streams, no TCP HOL
TLS	Optional → facto required	TLS required	TLS 1.3 built in
Interview focus	Keep-alive, caching headers	Multiplexing, HPACK idea	Why UDP, per-stream isolation

Caching headers (often paired with HTTP questions)

Cache-Control (max-age, no-store), ETag / If-None-Match for validation, Vary for cache key dimensions. Senior signal: distinguish browser cache vs CDN edge cache vs application cache.