Replacing the module crypto

Hi,

I’m opening this issue to know if you are open to replacing the use of crypto for more browser-friendly libraries like create-hmac and randombytes.

I’ve had a hard time using this library with Rollup, which led me to bundling my own version here: https://github.com/ethereum/js-ethereum-cryptography/tree/master/hdkey-without-crypto-config

I’d love not to have to do that, and the needed change is really small. I can prepare a PR for this, making sure that the builtin implementations are used on node.

[alcuadrado]

I don’t see how tiny-secp256k1 compilation step is related to browsers, since we fallback to pure JS when packaging for browsers. If the compilation fails you will use pure JS in node environment too.

I’m leading an effort to remove native dependencies (except N-API based ones) from the Ethereum js ecosystem. Why?

1. They are degrade the development experience by being slow to install.
2. They tend to break with new Node.js major versions
3. Many people, especially Windows users, don’t have a proper setup to use them. This is incredibly common in educational contexts (e.g. a workshop at a conference). In those cases, many people just desist.

As part of this process, I’m also making sure that the dependencies being used work well with web bundlers and don’t produce unnecessarily large bundles.

The problems with the crypto module are:

1. It is normally be replaced by crypto-browserify, which is HUGE.
2. It doesn’t work with Rollup.

Why hdkey and not bip32? hdkey is more popular in the Ethereum ecosystem, so migrating to a portable version of it is easier than migrating to bip32. This is especially important because this effort requires the buy-in of multiple open source projects.

[alcuadrado]

Thanks for the heads up, @junderw!