Hacker Newsnew | past | comments | ask | show | jobs | submit | Zedmor's commentslogin

Interesting approach modeling after Go's net/http but in Rust. The handler signature looks clean. Curious about the 2-3x performance claim — is that comparing against Go's standard library or something like actix/axum? Would love to see the benchmarks.

Hi thanks! My focus was developer ergonomics so it's rewarding to hear some of that has translated across.

The benchmark claim is based on a naive, simple "hello world" GET req/response, comparing the Go stdlib vs the equivalent handler using uhttp. I used oha to measure.

https://github.com/alshdavid/uhttp/blob/main/benchmarks/go/s...

vs

https://github.com/alshdavid/uhttp/blob/main/examples/basic/...

Output: https://github.com/alshdavid/uhttp/blob/main/crates/uhttp/RE...

Often a meaningless benchmark, but it does serve to measure the overhead from the http library itself. Happy to hear suggestions on the benchmark approach as I don't want to be misleading


I remember reading this when it first circulated. Still holds up — the clarity of his reasoning is remarkable.

Never heard of single one. I bet you just invented it.


I don't get it. If I spend some time reading through source code in github and now produce code, how come I can be liable for that? Why we hold humans and machines to the different standards? My brain runs on neural networks too and I can learn things from reading/observing. What is the difference?


Are those essays GPT-3 generated?


Funny but he seems to be old enough to realize that success in life and career is about people, not machines.


because your brain runs on same CNN software i.e. doing pattern matching, not image matching.


Kit is not killed, it just open sourced.


Read you resume. Un-fuckin-real. wow.


Geez. You weren't exaggerating.


Guidelines | FAQ | Lists | API | Security | Legal | Apply to YC | Contact

Search: