Big confusion around resolver.getAddress

[efstajas]

Just went through an extended period of confusion while implementing resolution of ENS names on L2s in my app.

My goal: If currently connected to an L2, resolve names on ETH mainnet. If there's an explicit address record for the L2 in question, use this. If not, fall back to the address for ETH mainnet, but only if that address is not a contract.

What I discovered the status quo to be:

• resolver.getAddress() returns the address for the name on Eth Mainnet. Makes sense.
• resolver.getAddress(10) (10 being the chainId for OP mainnet) returns the address for OP mainnet set on the record. Cool, so I can just pass chainId here.
• resolver.getAddress(1) (1 being the chainId for Eth Mainnet) however returns nothing, even though getAddress(), which defaults to mainnet, does. Huh?

I was really confused by this until I learned what a coinType actually is, the fact that ENSIP-11 defines a way to derive "legacy" coinTypes from evm chainIds, and that Ethers implicitly treats the coinType argument as a chainId and converts it to a coinType if it falls outside the range covered by ENSIP-11 and is not 60.

This has some IMO pretty odd consequences for the API behavior:

• As mentioned before, while resolver.getAddress(10) resolves to the OP Mainnet address, resolver.getAddress(1) does not resolve to ETH Mainnet address
• resolver.getAddress(60) resolves to the Eth Mainnet address, but 60 is at the same time the chainId for GoChain. If you're not aware of what exactly coinType is but noticed that passing chainId seems to just work, you would 100% expect to receive the address for GoChain when passing 60.
• The argument being called coinType implies that e.g. getAddress(10) would return the non-EVM address set on the record for Blackcoin, given that a coinType of 10 refers to Blackcoin as per SLIP-0044.

I think that ideally resolver.getAddress would just accept chainId only. I see how changing the behavior of this function now would be tricky though, so maybe deprecating resolver.getAddress and replacing it with two explicit functions getAddressForChain(chainId) and getAddressForCoinType(coinType) would get rid of the implicit conversion and resulting confusion.