EIP-6912: Versioned TokenIds for Dynamic NFTs

eip: 6912
title: Versioned TokenId Standard for Dynamic NFTs
author: James Wenzel (emo.eth, @emo_eth)
discussions-to: https://github.com/ethereum/EIPs/pull/6912
status: Draft
type: Standards Track
category: ERC
created: 2023-04-19
requires: 721

This EIP proposes an extension to the EIP-721 non-fungible token standard by introducing a Versioned TokenId standard for “dynamic” NFTs with on or offchain properties that may change over time. The new versionedTokenId is meant to track both the “identifier” of a token as well as its current “version” so that old outstanding orders and approvals for updated tokens are automatically invalidated.

See linked PR for latest details.

Hey @emo. Interesting EIP draft.

The described approach will probably work for marketplaces but fail for any protocol that uses escrow, like lending & renting with collateral. Those protocols assume that the tokenId will be immutable during the time locked in the protocol. By allowing tokenId updates (e.g., expired condition based on block.timestamp), those protocols cannot transfer the locked collateral back to the borrower or let the lender claim it, as they will not have the correct tokenId. The backward compatibility fails here.

I see another issue with deterministic tokenIds. If somebody updates the token to a new state and back, all existing bids will be invalidated.

Why not use EIP-5646?

Some non-editorial related comments:

Would you consider a function naming scheme closer to ERC-721? Something like currentVersionedTokenIdOf(...) and encodedTokenIdentifierOf(...).


I find your function names very confusing. I might recommend removing the separate concepts of “token id”, “versioned token id”, and “identifier”, and just using “token id” on its own?

That would mean you’d only need a single function: latestIdOf(uint256 tokenId).


You should mention in your Security Considerations something to the effect of:

If a token is locked in a contract (eg. as collateral for a loan), and the metadata changes causing a version bump, the token may be lost forever.


You don’t mention what happens to ERC-721-style approvals when a version number bumps in the Specification section. Do they get ported to the new token, or lost, or is it implementation defined?