05 · Grind week

Days 13–15 — Oct 19–20

Lost to the day job. On-call incident, three nights, zero project time. Logged for honesty, not for pity.

Day 16 — Oct 21

Back at it. The week’s goal: pin down the framing variants and the key schedule corner cases by hand so the protocol map is real, not vibes. REA carries a lot, but the nitty-gritty still resists — framing length prefixes change meaning between handshake and session, and two of the variants only show up in event bursts. Six hours in: one partial diagram. That’s it.

client ──hello(RSA pk)──▶ gateway
   ◀──session key(AES)──
   ├─── heartbeat ──▶   (cadence: ~15s, jittered)
   ◀── event push ───   (burst on rally/state change)
   ├─── cmd frame ──▶   [len][seq][cmd][payload]      ← variant B: [seq][len]…
                          ^ unclear when B applies

Day 17 — Oct 22

More hours, fewer results. The variant-B question is still open and I refuse to guess. REA’s best hypothesis needs a capture I can’t cleanly make yet. Logged three candidate triggers, no conclusion.

Day 18 — Oct 23

Called the week honestly in the log: long hours, thin results, one diagram. This is the phase of every RE project where the curve goes flat and the only sin is pretending it didn’t. Weekend plans: stop grinding the framing question directly, go read what the community already knows. Sometimes the fastest path through a wall is a door somebody else already opened.