Fixpoint problem: a block cannot contain its own CID, because computing that CID would mean finding a hash preimage. For the same reason, a cycle of CIDs has no encoding order. This is why schema references are names. References and naming walks through it.

See also

References and naming: the full argument and the naming layer.

hm:// URL: the names schemas use instead.

Self-description: the meta-schema's reference to itself.

Canonical encoding: why a CID depends on exact bytes.