reticiulum-specification/flows/README.md
Rob 5574d3bed3 flows: add lxmf-outbound-retry — process_outbound retry loop + state machine
Documents the outbound retry layer that wraps the existing per-method
send-* flows. Pinned to LXMF 0.9.7 / RNS 1.2.4 with literal-quoted
upstream source for every claim:

- 4-second tick cadence (PROCESSING_INTERVAL × JOB_OUTBOUND_INTERVAL)
- All seven retry constants (MAX_DELIVERY_ATTEMPTS, DELIVERY_RETRY_WAIT,
  PATH_REQUEST_WAIT, MAX_PATHLESS_TRIES, MESSAGE_EXPIRY,
  LINK_MAX_INACTIVITY, P_LINK_MAX_INACTIVITY) at LXMRouter.py:30-38
- Eight-state machine (GENERATING/OUTBOUND/SENDING/SENT/DELIVERED/
  REJECTED/CANCELLED/FAILED) at LXMessage.py:13-22
- The four terminal-state branches at top of process_outbound (lines
  2517-2558) and the three per-method retry branches (OPPORTUNISTIC
  2566-2592, DIRECT 2596-2673, PROPAGATED 2677-2730)
- fail_message semantics at LXMRouter.py:2395-2402

Includes a "what does NOT happen" section calling out common
misconceptions: no automatic DIRECT→PROPAGATED fallback, no
exponential backoff, no in-router persistence of pending_outbound,
MESSAGE_EXPIRY governs the propagation-node store not per-sender
retries, SENT is the terminal success state for PROPAGATED (not
DELIVERED).

No verifier needed per agent.md §1 — all claims are direct upstream
source citations.

Co-Authored-By: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
2026-05-09 10:09:02 -04:00

3 KiB

Flows

End-to-end chronological narratives for common Reticulum operations. Where SPEC.md is organized by layer (identity, header, token crypto, announce, LXMF, link, transport, framing), the documents here are organized by operation and walk through what each layer contributes in order — app-call → wire bytes.

The two views are complementary: SPEC.md tells you what each piece looks like; the flows tell you when each piece runs and what calls what. A flow document should not introduce new normative claims — every byte-level detail should be a cross-reference to the relevant SPEC.md section. If you find yourself describing wire bytes here that aren't in SPEC.md, that's a sign the spec has a gap to fill.

Status

Flow Status
send-opportunistic-lxmf.md
receive-opportunistic-lxmf.md
send-link-lxmf.md (DIRECT method, over a Reticulum Link)
receive-announce.md
send-resource.md (Resource fragmentation over a Link)
path-discovery.md (path? request, path-response wire detail, path-table population)
receive-resource.md (inverse of send-resource: ADV ingestion, part assembly, proof emission)
receive-link-lxmf.md (inverse of send-link-lxmf, including responder side of the handshake)
send-announce.md (build, sign, transmit, ratchet rotation, periodic re-announce)
forward-announce.md (transport-node rebroadcast logic, announce_cap, queue)
send-propagated-lxmf.md (PROPAGATED method, via a propagation node)
receive-propagated-lxmf.md (recipient pulling messages via /get)
lxmf-outbound-retry.md (process_outbound retry loop, per-message state machine, fail_message)

Conventions

  • Each flow targets one specific upstream operation. send-opportunistic-lxmf.md documents what LXMRouter.handle_outbound(lxm) does for an opportunistic message; it does not also cover Link or propagation paths — those get their own docs so the chronology stays linear.
  • Numbered steps are chronological. Each step that produces wire bytes cross-references the SPEC.md section that defines those bytes.
  • Source citations use the standard pip install rns lxmf install layout (RNS/, LXMF/) with file + line. Line numbers are pinned to the RNS / LXMF version named at the top of each flow; out-of-date line numbers should be fixed in a PR.
  • "Verified" claims must be backed by a tools/ script per ../agent.md §1. Flow docs inherit the verification status of the SPEC.md sections they reference — if a flow step relies on an unverified SPEC.md callout, the flow should mark that step as inheriting the unverified status rather than silently treat it as fact.