ADR-0059: Click-to-sign under the Electronic Commerce Act 2006 — no PKI, no third-party provider

Status

Accepted (2026-08-26)

Tags

rental, legal, e-signature, security, evidence, pdpa, product-boundary, adr-0058, adr-0055

Decision

Both parties sign a tenancy agreement inside the platform by typing their full name and NRIC, ticking agree, and submitting.
  • Evidence recorded per signature (rental_agreement_signatures): signer_user_id, signer_role, the full_name and nric as attested, document_sha256, signed_at, ip, and user_agent.
  • Double-signing is stopped by the database. A unique index on (agreement_id, signer_user_id) is the whole guard — not application logic.
  • Typed values are checked against the frozen snapshot. NRIC compares on digits only; the name compares normalized (trim, collapse whitespace, uppercase). What is stored is exactly what the signer typed, because that is what they attested.
  • The expected signer set comes from the agreement row — its landlord_user_id and tenant_user_id — not from a hardcoded count of two.
  • IP and user-agent stay in the database. The composite PDF’s signature page prints names, attested NRIC, timestamps with timezone, and the document hash. It does not print IP addresses.
  • Signature rows survive a void. If an agreement is voided after someone has signed, the signature rows remain as evidence, void_reason becomes mandatory, and both parties are emailed naming who voided and why.

Why

Malaysia has three plausible levels of electronic signature, and they are very far apart in cost. The Electronic Commerce Act 2006 gives legal effect to electronic signatures generally and does not exclude tenancy agreements from its scope. A typed attestation plus an evidence trail is therefore sufficient for the ordinary tenancy this platform serves. What makes the trail persuasive is not the look of the signature but the ability to show what document was signed, by which account, and when — which is exactly what document_sha256 + signer_user_id + signed_at record. A certified digital signature under the Digital Signature Act 1997 is stronger evidence, but it requires every party to hold a certificate from a licensed CA. For a DIY rental platform whose entire premise is removing friction between two ordinary people, that requirement would end sign-up before it starts. Keeping the IP out of the printed document is a PDPA 2010 judgement, not a legal-weight one. The evidence lives in the database either way; printing it hands each party a piece of the other’s network identity in a file they may forward to an agent, a family member, or a group chat.

Considered options

  • Drawn signature images — rejected. Familiar and reassuring, but adds no legal weight over a typed attestation while costing real frontend work, image storage, and a mobile drawing surface.
  • Certified digital signatures (Digital Signature Act 1997) — rejected. Strongest evidence available, but requires a CA certificate per party. This is the upgrade path if the platform ever serves commercial tenancies where the stakes justify the friction.
  • A third-party provider (DocuSign, Zoho Sign) — rejected. Fastest to build and strongest audit trail out of the box, but charges per tenancy on a product whose whole pitch is being cheaper than an agent, and moves the document flow off the platform — which also puts both parties’ NRIC into a third party’s system.
  • Application-level double-sign checks — rejected in favour of the unique index. A read-then-write check loses the race when both signatures arrive together, and this is a path where losing it means a duplicate completion.

Consequences

  • Identity is self-declared. A tenant can type any NRIC; nothing verifies it against a government record. Accepted at this volume, and identity verification is the first thing to add if fraud appears. This is stated in the spec’s risk table, not hidden.
  • The signature evidence is only as good as the account. Compromising a Better Auth session is enough to sign. This is the same trust level the rest of the platform runs on.
  • Adding joint tenants later needs no migration. Signatures are a separate table keyed (agreement_id, signer_user_id) with a signer_role, and completion counts against a derived expected-signer set. Only that derivation changes.
  • Agreements are not stamped. Validity between the parties is not the same as admissibility in court, which requires stamping with LHDN. The platform states this plainly on screen at issue and on the completed agreement.