Skip to content
ADR-0003Architecture decisionSupersededversion 1.0.030 October 2026

Suspended — research direction (2026-10-30, ADR-0072). The five hooks of this ADR are no longer binding; the network issues no token and handles no payment or value transfer. The text is kept as research work; the conditions for reopening are in ADR-0072 K3.

ADR-0003 — Value layer hooks ​

Status: Accepted (2026-08-06) — suspended 2026-10-30 (ADR-0072) Date: 2026-08-06 Decided by: Tamga Network project management Source: docs/tamga-network-varlik-katmani-strateji.md §4 + docs/tamga-network-cuzdan-odeme-agent.md §2, §4 (strategic input drafts). Strategy rationale → PM-AUTH-0001.


Context ​

The decision in PM-AUTH-0001: Tamga is an authorisation layer, not a settlement layer, and no payment feature is built in the short term. Being able to open the value/payment/agent door later, however, depends on a few architecture decisions taken today. What they have in common: almost free today, practically impossible later (the whole user base would have to migrate to new accounts). They therefore have to be fixed before any contracts/ Solidity is written.

This ADR builds on ADR-0001 (Besu/EVM) and ADR-0002 (sovereignty-first governance); the value layer is independent of the X.509 vs DID decision (DECISIONS D-ID-1) (layer A = an EVM address exists in either case).


Decision ​

Decision 1 — Chain accounts are smart contract wallets from the start (not EOAs) ​

A citizen's or institution's chain account must be a contract account (account abstraction, ERC-4337 style), not a plain key-to-address mapping (EOA) — even if today it only performs identity operations.

This single decision makes the following possible later: recovery, spending limits, agent delegation, gas sponsorship, multi-signature, policy rules. Gas sponsorship is also mandatory for a public service — citizens must not have to acquire a "gas token". Starting with EOAs, none of this can be added later.

Cost today: deploying one contract. ~zero.

Decision 2 — Strictly separate key domains + separate recovery ​

Three derivation paths, three lifecycles:

Identity key → credential presentation, pseudonym derivation.  Not recoverable*. Not delegable.
Asset key    → chain transactions, value transfer.             Recoverable. Delegable within limits.
Agent key    → transactions within a delegation.               Time-limited. Revocable instantly.

Recovery separation (binding): the recovery mechanism applies only to the asset key/account. The identity key and credentialsA digital document signed by an issuer and held in the person's wallet; the person shows only the fields that are asked for. are never part of it — otherwise the "recoverers" could take over the identity. (*Identities/pseudonymsA stable account identifier the wallet derives separately for each website; two sites cannot link the same person. are re-derived deterministically from the master seed; credentials are requested again from the issuerThe institution that signs and issues credentials: a university, a professional body, a public institution or a company. — this is re-derivation or re-issuance, not "recovery".)

Recoverer composition (decided 2026-08-06): asset wallet recoverers are user-chosen (family/trusted persons + optionally an institution), with an M-of-N threshold and a time lock. A mandatory state institution is NOT a recoverer — that would bring exactly the censorship/leverage paradox avoided in PM-AUTH-0001 back in through money. Asset recovery is a mechanism separate from and independent of identity escrow (SPEC-BC-0002 guardians). Concrete M-of-N, time-lock duration and abuse scenarios → PM-ID-0003.

Cost today: three derivation paths in the wallet. ~one day of work.

Decision 3 — The delegation registry is scope-generic ​

The agent delegation contract is written today, but scopes stay outside payments (logistics:verify, health:read). The scope field is designed to be extensible rather than a fixed string; adding a pay:* scope later becomes a one-line change. Writing a payment-specific delegation system later would invalidate every existing delegation.

Invariant principles of delegation (detailed in SPEC-AGENT-0001): an agent has no identity of its own, only derived authority; no indefinite delegation (validUntil mandatory); instant revocation (kill switch) without conditions; responsibility lies with the principal; every transaction is logged with its delegation reference; an agent cannot present identity credentials (transactional authority only).

Decision 4 — Credential gating as a generic primitive ​

The rule "this transaction requires that credential" should be a general contract library, not a payment-specific one:

solidity
modifier requiresCredential(bytes32 credentialType, bytes calldata proof) { ... }

Today in document verification; tomorrow in a rule such as "only KYC'd accounts may make this transfer" — the same primitive works. It is aligned with the SPEC-BC-0001 StatusList/IssuerRegistry checks.

Narrowed (2026-09-10, SPEC-AGENT-0001): this decision was narrowed. Since ADR-0008 moved the status list off-chain, the chain cannot see whether a credential is valid; therefore there is no on-chain credential gating in phase 0 and requiresCredential is not written as a chain primitive. Verification is done off-chain (SPEC-API-0001). The path to open it in the future (a signed statement by a registered verifierThe party that checks a presented credential: signature, issuer in the trust list, status and policy. Also called relying party. + freshness + nonceA random value used once; the verifier sends it and the wallet signs it, so an old presentation cannot be replayed.) is recorded as a design in SPEC-AGENT-0001 §4. Decision 3 (delegation) is unaffected.

Decision 5 — Do not issue an own token, but do not forbid token standards ​

Governance and technical architecture state explicitly that Tamga has no currency of its own (this sets up the relationship with central banks correctly from the start). But ERC-20 deployment is not forbidden at chain level — a central bank or licensed bank may later want to issue tokenised deposits here; Tamga is then the host, not the issuer.

Constraint: the right to issue tokens is tied to an allow-list — only institutions in the FINANCE category of the Issuer Registry, authorised by the state concerned (ADR-0002 onlyOwnerState + SPEC-BC-0001 IssuerCategory). This prevents unauthorised token spam and gives regulators confidence.


Rationale ​

  • Asymmetry of irreversibility: the five decisions cost ~zero today; taking them later requires user migration. The asymmetry clearly says "take them now".
  • Consistency with the strategy: the hooks are not features; none of them opens a payment capability today. They do not increase the risk of losing focus (PM-AUTH-0001).
  • Consistency with sovereignty: decision 2 (user-chosen recovery) and decision 5 (FINANCE allow-list) keep the censorship paradox away from money.

Consequences (binding for contracts/) ​

  1. The contracts/ wallet is designed as a contract account (decision 1); an EOA assumption is not accepted.
  2. The wallet/SDK produces three derivation paths (decision 2); recovery only on the asset path.
  3. The delegation contract makes scope extensible (decision 3).
  4. requiresCredential lives as a shared library in shared/ or contracts/lib (decision 4).
  5. No native token; ERC-20 deployment is gated by the FINANCE-issuer allow-list (decision 5).

Relations ​

  • PM-AUTH-0001 — the strategy rationale for this ADR.
  • ADR-0001 — Besu/EVM (layer A account model).
  • ADR-0002 — sovereignty-first governance (FINANCE permission, onlyOwnerState).
  • SPEC-BC-0001 — Issuer Registry / IssuerCategory / StatusList (data source for credential gating).
  • SPEC-BC-0002 — accountable disclosure (identity escrow, separate from asset recovery).
  • PM-ID-0003 (planned) — details of asset wallet recovery composition.
  • SPEC-AGENT-0001 — agent delegation surface (decision 3) + narrowing of credential gating (decision 4).

The five hook decisions were fixed on 2026-08-06. The contracts/ implementation follows these constraints. Recovery composition and agent delegation details were referred to separate documents. 2026-09-10: decision 4 was narrowed by SPEC-AGENT-0001 (no on-chain credential gating in phase 0; see the note under decision 4).