[RFC] NSEC Proofs as a Permissionless Release Trigger for TLDs

Context

Thus far, we’ve explored a design for claiming TLDs, originally as an RFC 'fast path..’ and later a v1 of the oracle design, since re-targeted to ENSv2 reserved entries, with a recently included deprecation policy archival tool for guarding against bitrot.

But we’ve yet to take a position on what happens at turn-down: operator failure, discontinuation, EBERO, moribund TLDs, etc.

Release

Under §4.2 of the initial TSG report: “the controlling entity MUST be the registry operator of the TLD.” In other words, Registry Operators (ROs) hold the token, lest ENS be non-conformant.

We’re aware that the report is in draft status and is therefore non-binding, but building towards it is cheap insurance.

Claims are current until expiry. A DNS-side termination has no onchain effect, so the TLD keeps resolving under an operator ICANN has already removed, for up to the full term (TBD).

Bringing it back under DAO custody means a register call, currently a DAO-gated action.

When ICANN pulls a TLD’s NS and DS from the root, we propose NSEC proofs as the authoritative signal for permissionless release; expiry is the fallback (more below).

How it works

Just as we proposed DNSSEC proofs as a permissionless way to transfer a TLD Node from the DAO’s safe to a valid RO, we can use NSEC proofs to end that claim, so the node can return.

// release: proven absence
release(tld, nsecProof):
    predicate.requireNoDelegation(tld, nsecProof) // no NS in the root
    registry.unregister(tld)

To avoid pedantic posting, check the predicate at steg-eth/tld-oracle-v2/nsec, with tests run against the published root zone.

Special Case: EBERO

We’ve not yet gone as far as to spec out what would happen if a TLD enters EBERO — for good reason: only 7 are in EBERO today.

A first suggestion would be to ship an NSEC3 shaped predicate for the TLD’s own zone, since an emergency operator has no reason to keep publishing the old RO’s apex record.

What we’re seeking feedback on

  • Should the design’s term align with existing ICANN convention (10 years)?
  • The oracle needs two roles granted for permissionless claims and releases, root ROLE_REGISTRAR and entry-scoped UNREGISTER, any push back?
  • Does an NSEC proof sufficiently demonstrate that a TLD’s delegation has been removed from the root? Or should there be a higher standard?
1 Like