41.06
item · Working in the open / Open source

Stellar Intents

experiment

We’ve learned a lot about integrating servers with each other: messages, outboxes, retries, sagas. A single-page app still mostly behaves like an RPC client, making synchronous HTTP calls and hoping. This project takes the distributed-systems view seriously at the edge. The browser works from memories (what it last read), makes guesses (what the screen shows before anything is decided), and sometimes owes the person an apology, in Pat Helland’s phrase.

The person asks; they don’t command. A command completes or fails. An intent is received, then decided, and when the answer is no, it comes with what can be done next.

Almost nothing on the wire is invented: it’s HTTP and Server-Sent Events, with the conventions written down. The server side is built on Wolverine and Marten (see the Critter Stack workshop), with a reference client in Angular.

It’s a work in progress, and the repository is clear about which parts are built and which are still guesses.

github.com/hypertheory-labs/stellar-intents