Canonical name resolver

ens-improvement-proposals/ensip-xx-canonical-name-resolver.md

@@ -0,0 +1,51 @@
+---
+description: Allows names to set a cname to alias records to another name
+---
+
+# ENSIP-XX: CNAME resolver
+
+| **Author**    | Jeff Lau \<jeff@ens.domains>
+| ------------- | ---------------------------------------------------------------- |
+| **Status**    | Draft                                                            |
+| **Submitted** | 2023-09-25                                                       |
+
+### Abstract
+
+Canonical Name records have existed for DNS since the 1987 in [RFC 1034](https://datatracker.ietf.org/doc/html/rfc1034). They allow aliasing a name to another name e.g. vitalik.eth to records.vitalik.eth. This specification outlines a resolver standard to implement CNAMEs at a resolver level without having to change the base ENS name resolution protocol.
+
+### Specification
+
+If a cname is found on a name, all records that are unset will be expected to proxy through the specified cname's records. Records that are already set on the resolver will ignore the cname.
+
+### Implementation
+
+```solidity
+interface ICnameResolver {
+    event CNameChanged(bytes32 indexed node, string name);
+    function cname(bytes32 node) external view returns (string memory);
+}
+```
+
+```solidity
+contract CnameResolver {
+    function addr(bytes32 node, uint256 coinType){
+        bytes memory currentAddr = versionable_addresses[recordVersions[node]][node][coinType];
+
+        if(currentAddr.length == 0 && cname.length != 0){
+            return ens.resolver(namehash(cname)).addr();
+        }
+
+       return currentAddr
+    }
+}
+```
+
+### Motivation
+
+* Allows separation of concerns between records and ownership of a high value name
+* Allows mirroring records for other names without having to set the records manually
+
+### Open Questions
+
+* DNS does not allow other records (except DNSSEC) when a cname has been set. ENS could optionally allow records on the main name to overwrite the cname, or it could choose to ignore them until a cname has been deleted. In the above specification, I've chosen to respect the main resolver's records.
+* We could also make the protocol more granular by allowing cnames per-record, which would allow you to delegate just a single record to a different name if necessary. This could be useful for allowing a service to take over that record for a name (e.g. contenthash)