# Conversation Log — Part V — Replay the Dynamic Idea Across Processes · Steps 35–38 (turns 63–67) > [⬅ Previous: Part IV](04-part-IV.md) · [Index](../conv.md) · [Next: Part VI ➡](06-part-VI.md) > > All parts: [Setup & Step 0](00-setup-and-step-0.md) · [Part I](01-part-I.md) · [Part II](02-part-II.md) · [Part III](03-part-III.md) · [Part IV](04-part-IV.md) · **Part V** · [Part VI](06-part-VI.md) · [Part VII](07-part-VII.md) --- ## [63] Student > for when its due part 5? **Tutor:** By the inferred mapping, Parts II–VIII all sit in the 22 Sep week, due 29 Sep, and Part IX was due 06 Oct — so a week behind. Gave the real dates instead (06 Oct "Now You Do It", 13 Oct Graded Exercise I) and a table of what each remaining part contributes, noting that Parts VI and VII are the ones the graded exercise leans on. ## [64] Student > finish reading 35 **Tutor:** Read Step 35 (lines 2180–2218): same dynamic idea across processes — one permanent control mailbox per service, `TOPIC` arrives, the service then watches that directory directly. Re-posed the question without answering it. ## [65] Student > now each service must be able to watch into others outbox if configured **Tutor:** Half right — the mediator already watched another service's outbox with the old helper. Corrected to the real constraint from Step 36: a service must watch its control inbox **and** a changing set of sources at once, which one blocking `watch(topic, receiver)` cannot do. That is why `ConfigurableMediator` accepted only one `ROUTE`. ## [66] Student > 1. the idea is manage access via key, register add a new folder to watch and create the key to access it, while key.cancel breaks the connection to a folder when we want it unsubscribe > 2. the queue has the advantage to keep the order of arrival **Tutor:** (1) ✅ with the detail that `FileSubscriptions` keeps both maps, `name → key` and `key → name`, so an event can be traced back to its mailbox. (2) Order is real but secondary; the reasons are that `observe()` must stay responsive for every directory, and that `record IncomingMessage(String mailbox, Message message)` lets one loop decide control-vs-data and change its own subscriptions. ## [67] Student > i have 5 ;-; (`find messageBoard-dynamic -name '*.msg' | wc -l` → 5) ## [67] Tutor — actions taken - Listed the five files with their types: two `TEXT hello` in `consoleIn/outbox` (one stranded from before `Configure.java` ran), one `TEXT uryyb` in `rot13/outbox`, and two `TOPIC` Intents in `rot13/inbox` and `consoleOut/inbox`. ## [67] Tutor — reply The count was right and the step's claim holds: **2 application publications** per flowing input (against 4 mediated), plus 2 one-time Intents and 1 line published before any subscription existed. Noted that `consoleOut/inbox` carries no payload at all now, and gave the ledger across Steps 18 / 27 / 38 (4 / 4 / 2 publications; 5 / 5 / 3 processes in the data path). Part V complete. --- > [⬅ Previous: Part IV](04-part-IV.md) · [Index](../conv.md) · [Next: Part VI ➡](06-part-VI.md) > > All parts: [Setup & Step 0](00-setup-and-step-0.md) · [Part I](01-part-I.md) · [Part II](02-part-II.md) · [Part III](03-part-III.md) · [Part IV](04-part-IV.md) · **Part V** · [Part VI](06-part-VI.md) · [Part VII](07-part-VII.md)