17 November 2026
From database to screen with Server-Sent Events
The data is in the database. Now someone wants to watch it change on a screen. There are three common ways to do it.
Polling, WebSockets or SSE
Polling: the browser asks "anything new?" every few seconds. It is simple, but most requests return nothing, and the screen is always a little late.
WebSockets: a two-way channel. Powerful, but you must handle the protocol upgrade, framing and your own reconnection logic, and a dashboard rarely needs the browser to talk back.
Server-Sent Events (SSE): one ordinary HTTP response that never ends. The server writes lines as events happen, and the browser reads them. It is one-way, which is exactly what a live view needs.
What SSE looks like
The response has the type text/event-stream, and each event is a few lines ended by a blank line:
id: 1042
event: reading
data: {"device":"greenhouse-3","value":21.4}
In the browser, new EventSource("/stream") is enough. If the connection drops, the browser reconnects by itself and sends
the id of the last event it received in a Last-Event-ID header. The server can then replay what was missed, which is
why the event identity from the ingestion brick pays off again here.
What the gateway has to get right
- Slow clients. A browser on a bad connection must not make the server buffer without limit. Drop it or skip to the latest value.
- Many clients, one source. Read the data once and fan it out, instead of one query per viewer.
- Authorisation. A stream is a long-lived request. Check who is asking when it opens, and decide what happens when their access is revoked while it is open.
Closing the loop
A reading published by a sensor, delivered by the broker, validated by ingestion, stored by the time-series database and pushed by the gateway reaches a screen in a fraction of a second. When one of those steps misbehaves, you will know where to look, because you built each one.