Abstract: An example embodiment of the present invention provides a process for inserting a source media stream into a target media stream, which media streams might be in container file format. The process opens and reads into memory structural information from a target media stream and structural information from a source media stream. The process then finds the nearest prior key frame to an insertion time and splits a chunk of media samples if the key frame's sample is not at the beginning of a chunk. The process adjusts the structural information for both streams in order to effect the mid-roll insertion of the source media stream into the target media stream. The process then writes onto an output media stream the adjusted structural information for the source and target media streams and then pipes onto the output media stream the media data for the source and target media streams.

Abstract: An example embodiment of the present invention provides a process for inserting a source media stream into a target media stream. The process opens and reads into memory structural information from a target media stream. The process then opens and reads into memory structural information from a source media stream. The process adjusts the structural information for both streams in order to effect the insertion of the source media stream into the target media stream. In the example embodiment, this adjustment avoids the use of temporary output storage. The process then writes onto an output media stream the adjusted structural information for the source and target media streams and then pipes onto the output media stream the media data for the source and target media streams. In particular embodiments, a container file format is used for the source media stream, target media stream, and output media stream.

Abstract: An example embodiment of the present invention provides a process for inserting a source media stream into a target media stream, which media streams might be in container file format. The process opens and reads into memory structural information from a target media stream and structural information from a source media stream. The process then finds the nearest prior key frame to an insertion time and splits a chunk of media samples if the key frame's sample is not at the beginning of a chunk. The process adjusts the structural information for both streams in order to effect the mid-roll insertion of the source media stream into the target media stream. The process then writes onto an output media stream the adjusted structural information for the source and target media streams and then pipes onto the output media stream the media data for the source and target media streams.

Abstract: An example embodiment of the present invention provides a process for inserting a source media stream into a target media stream, which media streams might be in container file format. The process opens and reads into memory structural information from a target media stream and structural information from a source media stream. The process then finds the nearest prior key frame to an insertion time and splits a chunk of media samples if the key frame's sample is not at the beginning of a chunk. The process adjusts the structural information for both streams in order to effect the mid-roll insertion of the source media stream into the tamer media stream. The process then writes onto an output media stream the adjusted structural information for the source and target media streams and then pipes onto the output media stream the media data for the source and target media streams.

Abstract: A centralized synchronization system synchronizes data stored in multiple devices distributed over a relatively unreliable network, to provide automatic recovery from synchronization failures. The synchronization system maintains a truth database indicating the true state of the data and action tables indicating actions to be performed on each of the devices during a next update of the data. Data is retrieved from the devices over the network, and actual states of the data are determined from the action tables and the retrieved data. The truth database and action tables are updated accordingly. An effective action table is then generated to account for any of the devices which were offline during the most-recent synchronization. The truth database and effective action tables are saved to a central database in an atomic transaction. The effective action tables are then used to update the data on the devices over the network.