The reader side already used Content-Length to know how many body bytes to pull; the writer side must provide it, or the client cannot tell where your body ends — especially once connections are reused and one response runs straight into the next. Computing it from len(body) guarantees it is always correct.

Derive it in the serializer rather than asking every handler to remember it. An empty body is Content-Length: 0, not a missing header — the count is always present, which is what keeps the framing unambiguous.

One knock-on effect: your responses now carry a second header, so the exact bytes you checked last lesson gain a Content-Length line — update that check, and remember headers serialize in sorted order (so Content-Length lands before Content-Type). Add it only when the caller has not set one, so a response that deliberately fixes its own Content-Length (a HEAD reply, later) is left alone.