[patrick-immometrica]

The blocking client sends a request body by reading it into an 8 KiB buffer and passing each chunk over a zero-capacity channel to the runtime thread. A large upload pays one thread handoff, one allocation and one zeroing per 8 KiB.

This raises the chunk size to 64 KiB, still capped by the body's length when that's known.

Uploading 256 MiB from Body::sized to a local server went from a median of about 1,170 ms (220 MiB/s) to 260 ms (980 MiB/s) over 7 runs.

The test posts a 200,000-byte body, sized and unsized, and checks that the server receives it byte-for-byte.

• The description, docs, and comments (words meant for humans) are written by a human, not by an LLM. (https://seanmonstar.com/micro/20260729-i-want-your-own-words/)