EIP 902: Token Validation
| Author | Brooklyn Zelenka, Tom Carchrae, Gleb Naumenko |
|---|---|
| Discussions-To | https://ethereum-magicians.org/t/update-on-erc902-validated-token/1639 |
| Status | Draft |
| Type | Standards Track |
| Category | ERC |
| Created | 2018-02-14 |
| Requires | 1066 |
Simple Summary
A protocol for services providing token ownership and transfer validation.
Abstract
This standard provides a registry contract method for authorizing token transfers. By nature, this covers both initially issuing tokens to users (ie: transfer from contract to owner), transferring tokens between users, and token spends.
Motivation
The tokenization of assets has wide application, not least of which is financial instruments such as securities and security tokens. Most jurisdictions have placed legal constraints on what may be traded, and who can hold such tokens which are regarded as securities. Broadly this includes KYC and AML validation, but may also include time-based spend limits, total volume of transactions, and so on.
Regulators and sanctioned third-party compliance agencies need some way to link off-chain compliance information such as identity and residency to an on-chain service. The application of this design is broader than legal regulation, encompassing all manner of business logic permissions for the creation, management, and trading of tokens.
Rather than each token maintaining its own whitelist (or other mechanism), it is preferable to share on-chain resources, rules, lists, and so on. There is also a desire to aggregate data and rules spread across multiple validators, or to apply complex behaviours (ex. switching logic, gates, state machines) to apply distributed data to an application.
Specification
TokenValidator
interface TokenValidator {
function check(
address _token,
address _subject
) public returns(byte statusCode)
function check(
address _token,
address _from,
address _to,
uint256 _amount
) public returns (byte statusCode)
}
Methods
check/2
function check(address _token, address _subject) public returns (byte _resultCode)
parameters
_token: the token under review_subject: the user or contract to checkreturns an ERC1066 status code
check/4
function check(address token, address from, address to, uint256 amount) public returns (byte resultCode)
parameters
_token: the token under review_from: in the case of a transfer, who is relinquishing token ownership_to: in the case of a transfer, who is accepting token ownership_amount: The number of tokens being transferredreturns an ERC1066 status code
ValidatedToken
interface ValidatedToken {
event Validation(
address indexed subject,
byte indexed result
)
event Validation(
address indexed from,
address indexed to,
uint256 value,
byte indexed statusCode
)
}
Events
Validation/2
event Validation(address indexed subject, byte indexed resultCode)
This event MUST be fired on return from a call to a TokenValidator.check/2.
parameters
subject: the user or contract that was checkedstatusCode: an ERC1066 status code
Validation/4
event Validation(
address indexed from,
address indexed to,
uint256 amount,
byte indexed statusCode
)
This event MUST be fired on return from a call to a TokenValidator.check/4.
parameters
from: in the case of a transfer, who is relinquishing token ownershipto: in the case of a transfer, who is accepting token ownershipamount: The number of tokens being transferredstatusCode: an ERC1066 status code
Rationale
This proposal includes a financial permissions system on top of any financial token. This design is not a general roles/permission system. In any system, the more you know about the context where a function will be called, the more powerful your function can be. By restricting ourselves to token transfers (ex. ERC20 or EIP-777), we can make assumptions about the use cases our validators will need to handle, and can make the API both small, useful, and extensible.
The events are fired by the calling token. Since Validators may aggregate or delegate to other Validators, it would generate a lot of useless events were it the
Validator’s responsibility. This is also the reason why we include the token in the call/4 arguments: a Validator cannot rely on msg.sender to determine the token that the call is concerning.
We have also seen a similar design from R-Token that uses an additional field: spender. While there are potential use cases for this, it’s not widely used enough to justify passing a dummy value along with every call. Instead, such a call would look more like this:
function approve(address spender, uint amount) public returns (bool success) {
if (validator.check(this, msg.sender, spender, amount) == okStatusCode) {
allowed[msg.sender][spender] = amount;
Approval(msg.sender, spender, amount);
return true;
} else {
return false;
}
}
A second check/2 function is also required, that is more general-purpose, and does not specify a transfer amount or recipient. This is intended for general checks, such as checking roles (admin, owner, &c), or if a user is on a simple whitelist.
We have left the decision to make associated Validator addresses public, private, or hardcoded up to the implementer. The proposed design does not include a centralized registry. It also does not include an interface for a Validated contract. A token may require one or many Validators for different purposes, requiring different validations for different, or just a single Validator. The potential use cases are too varied to provide a single unified set of methods. We have provided a set of example contracts here that may be inherited from for common use cases.
The status codes in the byte returns are unspecified. Any status code scheme may be used, though a general status code proposal is fortcoming.
By only defining the validation check, this standard is widely compatible with ERC-20, EIP-721, EIP-777, future token standards, centralized and decentralized exchanges, and so on.
Implementation
Copyright
Copyright and related rights waived via CC0.