Skip to main content
.com domains from $2.99 — free WHOIS privacy on every name

Transport, Measured

HTTP/2 vs HTTP/3 — why QUIC earns its keep on lossy networks

On clean fibre the two protocols look like twins. On a phone drifting between mast and Wi-Fi, HTTP/3's design decisions stop being academic — and that is where a growing share of your visitors live.

The short answer

Serve HTTP/3 if your stack offers it — on LiteSpeed it is native and there is no trade-off to weigh — but expect the gains where the protocol was designed to find them: high-latency routes, lossy connections, and mobile networks. There, QUIC's faster handshake, per-stream loss recovery and connection migration deliver improvements users feel; on low-loss wired connections the difference over HTTP/2 is real but modest.

The engineering summary: HTTP/2 multiplexed streams over a single TCP connection but inherited TCP's rule that one lost packet stalls everything behind it. HTTP/3 rebuilt the transport on QUIC over UDP, so streams lose packets independently, the handshake folds into TLS 1.3's, and a connection survives your phone changing networks. It fixes HTTP/2's last structural bottleneck rather than adding a new trick.

By the HostingFast team · Reviewed 25 August 2026

UDP

QUIC's foundation — delivery rules rebuilt per stream

1-RTT

New-connection setup; 0-RTT on return visits

0

Streams stalled by another stream's lost packet

alt-svc

The header that advertises HTTP/3 to browsers

Each HTTP revision has removed a queue. HTTP/1.1 queued requests per connection, so browsers opened six at a time. HTTP/2 multiplexed everything onto one connection and removed that queue — but underneath, TCP still promises in-order delivery of the whole byte stream, so a single lost packet holds up every stream sharing the pipe.

That residual queue is transport-level head-of-line blocking, and it is invisible in a lab with clean fibre and decisive on a train with two bars of signal. HTTP/3's answer was to stop asking TCP for permission and rebuild delivery semantics on UDP. This page measures what that rebuild is worth, and to whom.

What HTTP/2 fixed, and the queue it left behind

HTTP/2's multiplexing was a genuine advance: one connection, many parallel streams, header compression, no more sharding assets across hostnames. For its era's problem — request queueing — it was the right cure.

The remaining problem sits below it. TCP guarantees ordered delivery of one byte stream and cannot see HTTP/2's streams inside it, so packet loss anywhere stalls delivery everywhere on the connection until retransmission completes. The busier the multiplexing, the more innocent bystanders each lost packet takes hostage. On a 1% loss connection, that mechanism can make one connection perform worse than HTTP/1.1's six.

QUIC: streams that fail independently

QUIC implements reliability per stream, in user space, over UDP. A lost packet delays only the stream whose data it carried; your stylesheet does not wait for a retransmitted image chunk. TLS 1.3 is integrated rather than layered, so transport and encryption handshakes complete together — one round trip for a new connection, and zero for a resumed one.

The subtler gift is connection migration. QUIC identifies connections by ID, not by the IP-and-port tuple, so moving from Wi-Fi to mobile data carries the session across intact instead of timing out and reconnecting. For anyone browsing on the move, that is the difference between a pause and a page reload.

Where the gains concentrate — and where they don't

The wins scale with distance and loss. Long routes feel the handshake saving most, because every round trip costs real milliseconds. Lossy networks — mobile, hotel Wi-Fi, congested peering — feel per-stream recovery most. Mobile users switching networks feel migration. A UK visitor on office fibre reaching a London server will see numbers move only slightly.

This is why fleet-wide averages undersell HTTP/3: the improvement lives in the worst quartile of connections, which is precisely the quartile where visitors abandon. Judge it in your field data's p75 and p95 mobile numbers, not in a wired lab run.

Serving it, and proving it is on

On LiteSpeed platforms HTTP/3 is native — no modules, no sidecar, enabled as standard on every HostingFast plan. Apache and NGINX stacks assemble support with more ceremony, which in practice means many hosts still terminate at HTTP/2.

Verification takes a minute: check the response for an alt-svc header advertising h3, or watch the protocol column in your browser's network panel — the first request may ride HTTP/2 while the browser learns HTTP/3 is on offer, with subsequent requests switching. No site changes are required; the protocol negotiates itself.

A padlock standing in for the free SSL certificate on every plan

Modern transport as a default, not an upgrade

Because our platform serves through LiteSpeed, HTTP/3 is not a feature tier — it is simply how connections are answered when the browser asks. TLS 1.3, 0-RTT resumption and QUIC arrive together, on every plan.

Transport is the layer you buy by choosing a platform: no plugin provides it and no theme can compensate for its absence. It is worth one minute of header-checking before you commit to any host.

  • HTTP/3 served natively on every plan — no configuration, no tier gate
  • TLS 1.3 with session resumption, so return visitors handshake in zero round trips
  • Free SSL underneath it all, issued and renewed automatically
  • Verifiable in one request: look for h3 in the alt-svc header

Why HostingFast

Standard on every plan

Faster first contact

Transport and TLS handshake as one — a full round trip saved before the first byte, which is pure TTFB on long routes.

Grace under packet loss

Per-stream recovery keeps a lossy connection degrading gently instead of stalling wholesale — the mobile experience improves where it was worst.

Sessions that survive the commute

Connection migration carries the session across network changes, so leaving the house stops meaning starting over.

Quick Start

From order to online

  1. 1

    Check what you serve today

    One request to your homepage: an alt-svc header with h3 means HTTP/3 is on. Its absence means your visitors' worst connections are paying TCP's tax.

  2. 2

    Read your field data by network

    Core Web Vitals field data, segmented mobile against desktop. A wide gap is the signature of transport-sensitive traffic — the audience HTTP/3 helps.

  3. 3

    Move to a platform where it's default

    Transport ships with the server, not the site. On LiteSpeed it is native; migration is free, and the protocol column in devtools confirms the switch.

Built In

Loaded onto every plan

  • alt-svc header advertising h3 on your production domain, checked not assumed
  • TLS 1.3 in the handshake — QUIC requires it, and it is faster in its own right
  • Field data reviewed at p75/p95 mobile, where transport gains actually live
  • CDN and origin both speaking modern protocols, so the gain isn't lost at one hop
  • HTTP/2 retained as the negotiated fallback — some networks still block UDP
  • No site changes queued for this: transport upgrades belong to the platform

Frequently Asked

What people ask us most often

Do I need to change my website to use HTTP/3?

No. The protocol is negotiated between browser and server; your HTML, application code and plugins are oblivious to it. If the platform serves HTTP/3, capable browsers use it automatically and everyone else falls back to HTTP/2. The entire upgrade lives in the server — which is why it is a hosting decision.

Will HTTP/3 speed up my site on a fast broadband connection?

Modestly — mainly the round trip saved at connection setup. The protocol's structural advantages need adversity to show: latency, loss, network switching. Clean fibre to a nearby server gives it little to fix, which is exactly what the design predicts. Your mobile p95 is where to look for the difference.

Why was UDP the right foundation for QUIC?

Because TCP's delivery rules are fixed in operating systems and middleboxes that update on decade timescales. UDP provides bare datagrams, letting QUIC implement modern loss recovery, integrated encryption and connection IDs in user space — deployable by shipping software rather than by upgrading the internet.

Does HTTP/3 require a special SSL certificate?

No — the same certificate serves all HTTP versions. What QUIC does require is TLS 1.3, which is integrated into its handshake rather than layered on top. Any competent platform provides both together; on ours the certificate is issued and renewed automatically at no cost.

How can I check whether my site serves HTTP/3?

Request your homepage and look for an alt-svc response header containing h3 — that is the server advertising it. Or open your browser's network panel with the protocol column visible: h3 against your assets confirms it end to end. Note the first request often shows h2 while the browser learns h3 is available.

Keep reading

Changing hosts? Run through our checklist first.

A straightforward sequence for a switch your visitors never feel: which files move first, how to shift email across without losing a single message, the right moment to repoint DNS, and the two mistakes behind almost all the downtime we get asked to rescue.

You'll get the checklist email, then occasional pointers on keeping a site running fast. Unsubscribe the moment you want out — the privacy policy covers the rest.

Serve your worst connections better

HTTP/3, TLS 1.3 and 0-RTT resumption are the default on every HostingFast plan. Migrate free and watch the protocol column change.

View Hosting with HTTP/3 native plans