Summary of an article published on The Web Scraping Club on 2025-11-30 (free to read). Read the full article
Your browser and your HTTP client may not be speaking the same HTTP version to the same server. Most servers accept several versions, and the web is slowly moving to HTTP/3, which runs over QUIC on UDP instead of TCP. The question for scrapers is practical: do automated requests get anything out of it? Connection setup, behavior on unstable networks and rotating proxies, and possible effects on detection are the points where it could matter.
I start with what HTTP/3 is and how it compares with HTTP/2 and HTTP/1.1. Then I cover two ways to check whether a target supports it, using the HTTP/3 Check site and cURL. I look at which scraping clients can speak HTTP/3 and why browser automation with Playwright or Puppeteer does not use it by default. The hands-on part is a Python example with curl_cffi. I close with pros and cons, and an FAQ on fallback and HTTP/3 fingerprinting.
What’s inside
- Why only a minority of sites and a handful of clients support HTTP/3, and what that means for your choice of tooling.
- Why Playwright or Puppeteer may not use HTTP/3 even though the bundled browser supports it.
- What happens to curl_cffi’s HTTP version when you turn on Chrome impersonation.
- What the curl_cffi documentation says about HTTP/3 fingerprinting and why WAFs may inspect it less strictly than HTTP/2.
In the full article: The protocol comparison diagram and table, the cURL check on a real site, the list of HTTP/3-capable clients, the curl_cffi Python snippet with its output, and the full pros and cons list.