It’s true that the current proposal doesn’t convey that. In my mind there are two compelling reasons to go with the current solution:
- Doing otherwise would require changing every profile (record type) of every resolver in order to support a new parameter for wildcard domains. Even then, resolvers would only get
keccak256('label'), not the plaintext label. - Most uses for the hashed label can also be achieved with the namehash of the whole name. It’s easy to go from plaintext label and parent namehash to the namehash of the child node when looking things up.