Home page load, production builds, served locally over http:// — median of 7 cold-context runs in headless Chromium.
Bytes transferred, per stack
Number under each bar = separate HTTP requests needed
DOMContentLoaded, per stack (ms)
Time from navigation start to the document being ready — localhost, so absolute values are tiny
Read this carefully, it cuts against the obvious take: the framework stacks had
faster DOMContentLoaded despite shipping 2–8× more bytes, because bundling puts
everything in one JS file (3 requests total) instead of 6–7 small files (8–10 requests).
On localhost, request count matters more than byte count because latency is near zero. On
a real network with real latency per request, more requests cost more — but on a
real network with limited bandwidth, more bytes cost more. Neither chart on its own tells
you which stack is "faster" in production; it depends entirely on the network the site is
actually served over.