We develop a predictor approach to networked control with sampled-data and unknown time-varying transmission delays. First, we consider a state-feedback predictor. We show that an arbitrary small controller-to-actuators delay uncertainty may lead to a non-small residual error in networked control systems and reveal how to analyze such systems. Then we consider output-feedback control and design a sampled-data observer, which gives the state estimate used in a predictor. We emphasize the purely sampled-data nature of the measurements delays in the observer dynamics. This allows for an efficient analysis via the Wirtinger inequality, which is extended here to cope with exponential stability. To reduce the amount of sent measurements, we incorporate an event-triggering mechanism. Numerical examples illustrate that predictor-based control allows to increase essentially network-induced delays, whereas the event-triggering mechanism significantly reduces the network workload.