At what point you need computing performance on an I/O bound task?
The only performance that matters in http requests is async support. You can have 1000 pararel curl processes and will still be dusted by a single async python process.
That may be true, that has never matter to me. What has mattered however is portability. But it doesn't really matter what weaknesses there are, what matters is that they exist. Because they exist, HTTPie is not a strong of recommendation as ripgrep or fd.
You can think of it as a spectrum ranging from incomparable to strictly better. So you can't say that one should use Python over grep for example; you can't even compare the two because that doesn't make sense. They aren't even in the same space. Ripgrep however is a strict improvement over grep, it is faster, more featureful, safer, and has no downsides (with the exception of age, as I've mentioned).
I think HTTPie is very close to the end of the spectrum of being a strict improvement, but it's held back by various little things that don't matter to many but still matter.
The only performance that matters in http requests is async support. You can have 1000 pararel curl processes and will still be dusted by a single async python process.