drogna

What it is

A synthetic ocean, sensors that sample it, a forecast loop that assimilates what they report, and a query layer that serves the result through OGC API-EDR and SensorThings. All of it is a genuine program. All of it runs in one browser page.

The components do not know that. They talk to each other over a broker and read each other's data over HTTP, in the shapes those standards define — the difference is that the transport is a shim rather than a network. That boundary is the point of the exercise, and the reason the whole thing can be a URL rather than a deployment.

The name means nothing. It was coined so that it would carry no association with any domain, organisation or piece of work.

Run it

Open the demo — the current build of the default branch. Every visit provisions a fresh seeded run; nothing persists between visits.

Views are addressable, so a link can open the page at the thing being discussed rather than at the front door. The same page explains where the per-branch review instances live.

Where to start

• The demo — the running thing, and the instances published for review.
• The architecture — what the seam is, and what it buys.
• The component reference — every component, what it may publish, what it may hear. Generated from the same declaration the program is built from.
• The standards primers — SensorThings, OGC API-EDR, CF conventions and CoverageJSON, for readers who have not met them.
• The algorithm derivations — the mathematics that is not obvious from the code.
• The glossary — start here if the vocabulary is unfamiliar. Half of it is oceanographic and none of it is assumed.
• The decision records — why it is built the way it is, including the record that ended Version 1.

Version 2

Version 1 was twelve containers across Python, SQL and nginx, deployed to a server. It worked, and reviewing a change to it meant reasoning across containers, so it was slow. Version 2 keeps the architecture and moves it behind a wire-protocol seam into the browser; the reasoning is ADR-0027.

The Version 1 record is not deleted. Its blog entries, its eighteen-component reference and its architecture overview are kept in the archive, marked as what they are: an accurate description of software that no longer runs.

How this site is made

The source is markdown in the repository, built by a TypeScript script and pushed into the gh-pages estate. That branch is machine-owned: anything edited there by hand is overwritten by the next publication.

The site loads nothing from another origin — no fonts, no scripts, no stylesheets, no analytics, no tracking of any kind. It declines to be indexed. Outbound hyperlinks to standards documents are the only external references, and those are links a reader chooses to follow, not resources the page fetches.