spec: BRIDGE-1 §3.2 the session NAT covers an inbound default

[JarbasAl]

🤖 Auto-generated by Claude Fable 5.1 (claude-fable-5-1) via Claude Code — NOT human-reviewed. Verify before acting.

Identity mapping (NAT). Just as a bridge MAY rewrite
context.source on outbound messages ("topology hiding", §3.1), it
MAY also rewrite context.session.session_id as messages cross the
boundary — translating a participant-local session identifier into a
hub-side identifier and vice versa.

BRIDGE-1 §3.2 makes session-id translation a MAY, with no carve-out for the reserved value "default". SESSION-1 §3.1 calls a remote participant's use of "default" deliberate impersonation of the device-local session and leaves authorizing it to layer 2, but a bridge that skips the NAT for that one value lets an inbound message reach the orchestrator's default-session store directly (SESSION-2 §5.1), where an omitted field leaves the stored value unchanged and a present field replaces it. One inbound message can rewrite persistent device-local state, and a session stored that way can be replayed on a later, unrelated interaction.

The new text makes the NAT total: a bridge MUST map every inbound session_id, including "default", to a bridge-side identifier, unless a layer-2 grant outside the specification authorises the participant to act on the default session.

[coderabbitai]

Important

Draft PR not reviewed

Draft PRs are not automatically reviewed by default.

• Trigger a manual review

To automatically review draft PRs, update your CodeRabbit configuration:

reviews:
  auto_review:
    drafts: true

---

Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out.

❤️ Share

• X
• Mastodon
• Reddit
• LinkedIn

_{Comment @coderabbitai help to get the list of available commands.}