~ knotless Search/
↑↓ move↵ open⇧↵ all results

hum.delve.town/at_mcp

hum's fork of GroveResearch/at_mcp. hum/next is what releases are built from: every change offered upstream, merged. listeners is the listener work in progress (docs/listeners.md). Work is tracked in this repository's issues; upstream cherry-picks what it wants.

at_mcp 0.3.3-hum.4 Pre-release 1 package

0
0

Per-sender cap, and agents that wake each other

Open
Open hum.delve.town opened this issue 0 comments

Design: docs/listeners.md, "Who it serves", S10, and "The model", item 4.

Coordinating agents need "only these agents, only these record types, may wake me", and two agents must never wake each other forever. A flat write quota was a crude guard against the second. If the quota defaults to off (Write quota: let it be off), this is the guard that aims at the loop itself.

Do:

  • A listener may name which accounts can reach it (from).
  • A per-sender cap per listener per window, layered (operator ceiling > account default > listener value). Arrivals over it are kept as updates but don't ring, and are counted as capped.

Tests that would show it wrong

  1. With from: [A], an arrival from B is not an update.
  2. A sender over the cap doesn't ring; its arrivals are still readable and counted as capped. Another sender still rings.
  3. Two fake agents that reply to each other's replies as fast as they can: each agent's rings stay within its cap, and the operator view shows the capped count.
  4. An account default can't exceed the operator's ceiling (refused when saved).