Codd actually use a synthetic identifier for employee ID in an example in the original paper. So I don't think he is against synthetic identifiers in principle.

Codd actually use a synthetic identifier for employee ID in an example in the original paper. So I don't think he is against synthetic identifiers in principle. But if a natural key exist of course a synthetic key need not be introduced.