This paper introduces a new framework, that is, the virtual video (vvideo) capture device on Unix operating system. The vvideo capture device is suitable for realizing augmented reality applications as networked generic applications. It enables the augmented reality applications to become generic applications via the kernel layer so that all ordinary networked software for using videos can incorporate augmented reality applications directly and easily. Also, we discuss the basic mechanism of the announcement method on Internet which enables all sites to reach any augmented reality applications on Internet. Then, we propose a new video information management model for dealing with spatial and temporal characteristics as metadata of video data. Important components of the new model are Fragmented Reality (FR) and Recalled Reality (RR) for storing video data from the real world and for retrieving and processing video data dynamically. Finally, we show some augmented reality prototypes on Internet based on our new frameworks.