coordination without control
<!DOCTYPE html> <html lang="en"> <head> <meta charset="UTF-8"> <meta name="viewport" content="width=device-width, initial-scale=1.0"> <title>coordination without control / sotoalt</title> <meta name="description" content="how systems organize without central authority"> <link rel="stylesheet" href="../style.css"> <link rel="icon" href="data:image/svg+xml,<svg xmlns='http://www.w3.org/2000/svg' viewBox='0 0 100 100'><text y='.9em' font-size='90'>~</text></svg>"> </head> <body> <main> <nav class="breadcrumb"> <a href="../index.html">~</a> <span class="sep">/</span> <a href="../index.html#thoughts">thoughts</a> <span class="sep">/</span> <span class="current">coordination without control</span> </nav>
<article> <header class="post-header"> <h1>coordination without control</h1> <p class="meta">2024.11 · 5 min</p> </header>
<div class="content"> <p> Most systems assume coordination requires control. A central server. A master process. A coordinator node. But some of the most robust systems coordinate without any central authority. </p>
<p> Git is peer-to-peer. Bitcoin has no master node. Ant colonies have no CEO. These systems achieve coordination through <em>protocol</em>, not hierarchy. </p>
<figure> <img src="../images/agent-topology.svg" alt="Multi-agent topology diagram"> <figcaption>agents communicate through shared state, not direct calls</figcaption> </figure>
<h2>stigmergy revisited</h2>
<p> Ants coordinate by modifying their environment. One ant leaves a pheromone trail; another follows it; the trail gets stronger. No ant tells another what to do. The environment <strong>is</strong> the communication channel. </p>
<p> In software terms: the database is the message bus. You don't call other services — you write to shared state and let others react. </p>
<h2>WAWE and environmental coordination</h2>
<p> This is the model behind WAWE's agent coordination. Agents don't call each other. They read from and write to a shared SQLite database. File reservations, task queues, learnings — all stored in the environment. </p>
<p> When Agent A reserves a file, Agent B simply <em>sees</em> the reservation (by querying the database) and acts accordingly. No direct communication needed. If Agent A crashes, the reservation eventually expires. The system heals itself. </p>
<h2>the tradeoffs</h2>
<p> This approach isn't always better. You lose: </p>
<ul> <li>Immediate feedback (you poll instead of push)</li> <li>Strong consistency (eventual consistency is the norm)</li> <li>Debugging clarity (no call stack, just state changes)</li> </ul>
<p> But you gain: </p>
<ul> <li>Resilience (no single point of failure)</li> <li>Flexibility (add new agents without changing existing ones)</li> <li>Simplicity (each agent is independent)</li> </ul>
<p> Not every system should work this way. But when building systems that need to be robust and evolvable, it's worth considering: can we coordinate through environment rather than control? </p>
<hr>
<p class="footnote"> Further reading: <a href="https://en.wikipedia.org/wiki/Stigmergy">Stigmergy</a> on Wikipedia. Pierre-Paul Grassé coined the term in 1959 studying termites. </p> </div> </article>
<footer> <p class="nav-footer"> <a href="../index.html">← back</a> </p> </footer> </main> </body> </html>