Issues of representing context illustrated by video-surveillance applications