Which functions should be faster?

[paulmillr]

If your app has is bottlenecked because of scure-base, write a comment here.

The current architecture is cool and readable, but not the fastest one. For example, hex.decode is slower than in noble-hashes / noble-curves. hex decoding is not used anywhere here, but it is used in scure-btc-signer. base64 is also a bit slow, but so far has not been a bottleneck.

[AlexErrant]

Perhaps out of scope and outside your comfort area, but I wonder if wasm offers any worthy advantage. Might be that the interop cost isn't worth it for like, 512 bits though.

[paulmillr]

No wasm. Wasm creates opaque binaries that are hard to inspect and may contain malware.

[AlexErrant]

Sorry for going further off-topic, but doesn't that argument also apply to minfied JS? Devs typically inspect source code, not compiled/transpiled output. Also, you don't have to ship opaque binaries - there was some drama recently in the Rust ecosystem when Serde tried to do that.

[paulmillr]

doesn't that argument also apply to minified JS

yes. also to all build tools. My goal is to provide secure, audited / auditable, small packages. It's up to user to use additional build tools

you don't have to ship opaque binaries

What do you mean? If I publish package with wasm, it would be necessary to include wasm-compiled binaries themselves in the package. That's industry standard. We've (as NPM community) moved away from building packages on user machines because node-gyp shown that it's too unreliable.

[AlexErrant]

Ah derp you're right. I got lost in Rust land for the past few days and lost (or gained) some brain cells.

Now that I think about it... its kinda insane that NPM ships opaque binaries.

Javascript.

[paulmillr]

Sad!

[paulmillr]

30

[o-az]

are there any published benchmarks for this library? Looking for something like this https://www.xaymar.com/articles/2020/12/08/fastest-uint8array-to-hex-string-conversion-in-javascript/

[paulmillr]

npm run bench