feat(identity): one-call registry reads, a viem-free handle export, and the ENS Parent Name

[SupremaLex]

Five read functions on IdentityRegistry, a viem-free handle export in @libid/contracts, the ENS Parent Name as a constant, and a NatSpec pointer fix. This PR offers a menu: keep the reads that are worth keeping and drop the rest.

The reads

All five are on IdentityRegistry and IIdentityRegistry. Storage is unchanged; both storage-layout checks pass.

Function	Returns	Why
handleBindingOf(platformId, handle)	(holder, observedAt)	A holder and its proof age in one call, with resolveHandle's semantics. LibID (libID #71) needs this. Today it infers state from which registry call reverts.
idBindingOf(platformId, id)	(holder, observedAt)	resolveId plus the proof age.
handleOfId(platformId, id)	(handle, current)	An account's latest handle and whether it still holds it, without an indexer.
idOfHandle(platformId, handle)	id	Which account holds a handle now, so a client can see a handle change hands without remembering the previous id.
normalizeHandle(platformId, handle)	normalized	The handle under this chain's current rules, so clients (@libid/ens, the ENS gateway) can use the chain's rules instead of copies made at release time.

Semantics:

• Platform gate. The first four revert UnknownPlatform on an unwired platform, as resolveHandle and resolveId do. normalizeHandle follows handleHashOf: it answers for a platform with rules but no verifier, and reverts UnusableHandle(problem) on text the rules refuse.
• Refused text. In the other reads it answers (0, 0) or "", not a revert.
• A retired handle. handleBindingOf gives (0, old observedAt), as handleBinding does. That is the watermark a new proof has to beat.
• A handle its identity renamed away from. idOfHandle returns "" until another identity proves it, matching resolveHandle's zero.
• One code path each. resolveHandle and resolveId now go through handleBindingOf and idBindingOf. identitiesOf and handleOfId share one _handleOf. handleHashOf hashes normalizeHandle's result. No existing ABI changes.

TS

• @libid/contracts/handle: normalize, rulesFor, the rules, the handle vectors and ENS_PARENT_NAME, with no viem at load time. A test walks the subpath's imports and fails on any import that isn't relative, viem included. ./identity still needs viem. This lets @libid/ens (libID #109) drop its viem peer.
• ENS_PARENT_NAME = 'handles.link': also exported from ./identity and the root. The doc comment says the testnet deployment uses testnet.handles.link.
• Typed reads: every resolve.ts wrapper now takes its function name, arguments and return type from the generated identityRegistryAbi. A misspelled name or a wrong tuple fails tsc. The tests for the four new wrappers decode with the real ABI.
• No TS normalizeHandle wrapper: it would send the handle text to the RPC. The README points to normalize(handle, await rulesOf(...)), which runs locally. The contract view and its Rust binding stay.

For review

• Interface. The five reads live in IIdentityRegistry, which already carries reads (handleBinding, handleHashOf, …) and now describes itself as the registry's reads for contracts and clients alike. The escrow tests' SettableRegistry stubs revert ("not modelled"), so they can't give results the real registry never would.
• handleOfId.current after rules narrow. It stays true while resolveHandle, handleBindingOf and idOfHandle answer zero for that handle. That matches the existing identitiesOf.handleCurrent (pinned by test_handleCurrentReadsTheNodesNotTheRules) and is documented in the NatSpec, the Rust docs and the TS type.
• Names. idOfHandle is live: it answers only while the handle has a holder, unlike the raw idNodeByHandle pointer an indexer may mirror. The NatSpec says so.
• Bindings older than the preimage maps. On a proxy holding bindings written before idPreimages and handlePreimages existed, handleOfId and idOfHandle return empty for those bindings, the same gap identitiesOf has.
• Gas on existing views. Measured in one probe transaction against main: resolveId and resolveHandle cost about 110–150 gas more, because they run through the new reads, one code path kept on purpose. handleHashOf and handleNodeOf cost 44–88 more, and the five new selectors add about 22 to dispatch. identitiesOf costs about 74 more per item. Restoring main's exact loop body measured worse, so the cause looks like compiler output for the extra string-returning functions rather than this source. The committed verifier gas snapshots are unchanged.
• HandleResolver NatSpec. It cites libID's ENS integration spec (specs/ens-integration.md, REQ-ENS-KEY-01) without a PR number. It states the three-way key split, that the current deployment does not meet it yet, and that resolvers sharing a Signer must sit at different addresses (REQ-ENS-KEY-02).

Checks

• Solidity:
  • forge build, and forge test 673/673, including 22 new tests: bound, unbound, refused and unwired cases, rename and takeover, two identities of one holder, platforms kept apart, rules that narrow, case and @ variants, and agreement with identitiesOf.
  • forge fmt --check; both storage-layout scripts.
  • The regen-*.py --check scripts; gas snapshots unchanged.
• Rust: cargo clippy -D warnings, and cargo test --all including the binding drift test and the anvil tests. cargo +nightly fmt --check ran on a local nightly, not CI's pin.
• TS: install, codegen, build, typecheck, 43/43 tests including the viem-free guard, lint, fmt.

🤖 Generated with Claude Code

https://claude.ai/code/session_01Ar21wPcHzNprswTzLJd3MA