At many levels of computer system design there is a severe mismatch between ease of achieving bandwidth and latency goals. This paper suggests a new approach to achieving user-level latency goals, by focusing on streaming data as fast as possible and sampling the streamed data to satisfy individual requests. By analogy with transportation, the proposed approach is a mass-transit approach, by contrast with conventional strategies which are closer conceptually to a a single person in a car. Two examples are given: disk delay lines, modelled on one of the oldest kinds of computer memory, and a scalable video on demand (VoD) architecture.