polling vs long-polling vs push: ментальная модель

Question

Accepted Answer

Short-poll (timer), long-poll (server holds HTTP), push (WebSocket/SSE). Decision = latency budget vs infrastructure cost.

## Разбор

How client learns server changed — three shapes:

```
SHORT-POLL   C→S "anything?"  S→C "no"   (wait 5s)  C→S "anything?" S→C "yes, here"
             └ latency ≈ poll interval; N requests whether or not there's news

LONG-POLL    C→S "anything?" … S holds … (event!) … S→C "yes, here"  C→S "anything?" …
             └ latency ≈ 0; one request PER message; reconnect after each

PUSH         C→S open stream ─────────────▶  S→C msg  S→C msg  S→C msg …
             └ latency ≈ 0; one connection, server writes whenever it likes
```

Short-poll — simple, laggy. Long-poll — push-like latency on plain HTTP. Push — persistent pipe, you own lifecycle.

## Итог

Not «newest wins» — match pattern to event rate, infra budget, and who owns reconnect.