[Experimental]: Webcrypto Ed25519 Polyfill: should private key be able to be extractable?

[mcintyre94]

Overview

There's a mismatch between the test description and the code behaviour and I'm not sure which is correct. We need to either correct the test description, or change the code to match the test description behaviour. https://github.com/solana-labs/solana-web3.js/blob/5e1b2e5d451f4016b1d846d84b3fa3309e43e494/packages/webcrypto-ed25519-polyfill/src/__tests__/secrets-test.ts#L70-L76

• the test description says extractable should always be false
• the test itself actually checks that it matches the extractable value that is passed in

The code itself also uses the extractable value: https://github.com/solana-labs/solana-web3.js/blob/5e1b2e5d451f4016b1d846d84b3fa3309e43e494/packages/webcrypto-ed25519-polyfill/src/secrets.ts#L58-L63

While it explicitly overrides extractable for the public key: https://github.com/solana-labs/solana-web3.js/blob/5e1b2e5d451f4016b1d846d84b3fa3309e43e494/packages/webcrypto-ed25519-polyfill/src/secrets.ts#L64-L69

My guess is that we just need to fix the test description, but I figure it's worth making sure there's not a code bug here.

Also see https://github.com/solana-labs/solana-web3.js/pull/1681#discussion_r1355296887

[buffalojoec]

Seems like the behavior of the Subtle Crypto API is to allow either value for extractable for both, since you don't actually tell the API which is public or private when you import. I think we're probably nudging people toward marking public keys extractable since they need to be used to convert to a Base58EncodedAddress likely using this:

https://github.com/solana-labs/solana-web3.js/blob/5e1b2e5d451f4016b1d846d84b3fa3309e43e494/packages/addresses/src/public-key.ts#L5-L14

This function should fail if the key isn't marked as extractable.

[mikemaccana]

Also related - we may not actually need the polyfill as it's only for EoL node: https://github.com/solana-labs/solana-web3.js/issues/1680

[steveluscher]

…we may not actually need the polyfill as it's only for EoL node…

It's for browsers, mainly. And probably Hermes (React Native) although I haven't checked yet.

[steveluscher]

Honestly, I probably just got the test description wrong because I couldn't use the %s printf placeholder twice in the description string. https://jestjs.io/docs/api#1-testeachtablename-fn-timeout

[github-actions[bot]]

:tada: This issue has been resolved in version 1.87.2 :tada:

The release is available on:

• npm package (@latest dist-tag)
• GitHub release

Your semantic-release bot :package::rocket:

[github-actions[bot]]

Because there has been no activity on this issue for 7 days since it was closed, it has been automatically locked. Please open a new issue if it requires a follow up.