Implement Access Control

[martriay]

No description provided.

[andrew-fleming]

The Solidity implementation for AccessControl stores the roles as type bytes32. Cairo field elements, however, can only store 252 bits (31.5 bytes). For interoperability, it seems we cannot use a regular felt to store the role on StarkNet. One potential solution is to store the role as a mapping of two felts. @frangio @Amxx opinions on best way to proceed?

[milancermak]

@andrew-fleming What interoperability do you have in mind in this case?

[Amxx]

At first, the PR proposed (#239) used a uint256. If was felts that felt would be much easier to manipulate with than uint256.

IMO, using 252 bits for roles is way sufficient to avoid collisions. It has the drawback that you can't simply reuse the bytes32 roles transparently, but I don't think it's an issue. We can "just" take 252 bits of keccak256("PROPOSER_ROLE") and be good with it

[andrew-fleming]

@milancermak one specific example (that's fresh because I'm working on it) is with standard Timelock roles a la keccak256("TIMELOCK_ADMIN_ROLE"), keccak256("PROPOSER_ROLE"), etc.

[milancermak]

Right, but what's the benefit of having the role const the same in a Solidity contract and a Cairo contract?

[andrew-fleming]

We can "just" take 252 bits of keccak256("PROPOSER_ROLE") and be good with it

😆

That seems reasonable to me^

Right, but what's the benefit of having the role const the same in a Solidity contract and a Cairo contract?

@milancermak ah sorry, I didn't answer your question directly. I think the benefit is in easing the transition for existing applications following the standard. It'd be much simpler, IMO, for cross-chain applications that implement AccessControl to be able to use the same value (thus the same or similar mechanism to calculate said value). Though, @Amxx's solution might resolve this discrepancy with the best of both worlds approach

[Amxx]

Note, if you have an app that lives both in the EVM and Starknet spaces, you can stick to 252 bits roles and use them in both places.

[frangio]

I would definitely prioritize usability over strict compatibility with Solidity. What is the size of hashes in Cairo natively? If felt, I'd say let's use felt.

[martriay]

We can use felt, but what should we represent in them? Should it be the first or last 252bits of keccak256("ROLE") or should we use cairo's own hashing on a short string?

[frangio]

No strong opinion on this. Note that the format of roles is not mandated by AccessControl, it's up to the user and recommendations can change in the future. We've seen users of Solidity's AccessControl use various formats for their role ids including different hash functions or the direct string bytes.

[martriay]

We can maintain certain agnosticism then. Let's do felts.