Skip to content

Decide: signature replay boundary — position binding or defer #32

Closed

Description

@tonibergholm

Part of #24

Question

Signatures bind identity to a snapshot id, not a branch position, so a legitimately signed snapshot can be replayed elsewhere in history; merge/rebase/amend results start unsigned by design (P22 threat model; audit known-boundary finding). Does position/ancestry binding (or another mitigation) get specified now, or is this boundary re-affirmed and deferred with recorded rationale?

Activity

  1. self-assigned this
    on Jul 10, 2026
  2. tonibergholm commented on Jul 10, 2026

    @tonibergholm
    MemberAuthor

    Resolution

    Re-affirmed and deferred. No position-binding mechanism is specced in this map; P22's threat model stands, restated more sharply than the audit's phrasing.

    Sharpened boundary (recorded): a Snapshot's parents are part of the content its id hashes, so a P22 signature over the snapshot id already binds identity to the snapshot and its entire ancestry — a signed snapshot cannot be grafted under different parents without voiding the signature. The genuine residue is ref binding only: nothing stops a hostile remote from pointing a branch name at an older or side-branch legitimately-signed snapshot (rollback / ref-swap) — the same gap git addresses with signed pushes / gittuf-class ref metadata. "Replayed elsewhere in history" overstates the surface; "replayable to a ref tip within its own history" is the accurate claim, and docs/THREAT-MODEL.md should say it that way (folded into the assembly ticket's doc sweep).

    Defer rationale: the attack requires a hostile shared or multi-writer server; no such deployment exists (pre-1.0; P29's serve auth targets single-operator use). The correct fix — gittuf / push-certificate-shaped signed ref attestations with freshness/monotonicity semantics, verified at fetch/clone — is a trust-model and protocol design that deserves its own charted effort when server deployments are real, not a corner of this map.

    Revisit triggers: the first real multi-writer or hosted sc serve deployment, or hosting sc repos on untrusted infrastructure.

    Also kept as-is: P22's re-sign-after stance for merge/rebase/amend results (a new snapshot id is a new claim); threading --sign through history-op completions stays an un-promoted ROADMAP follow-on — it shrinks an ergonomic window, not this replay surface.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

Labels

wayfinder:grillingWayfinder ticket: HITL grilling/decision

Type

No type

Projects

No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions