A system and method for controlling video and/or audio streams between a client and server over a communications channel is disclosed that utilizes an application layer streaming communications protocol that includes features of current pull and push style application layer streaming protocols. Using the streaming protocol, client applications on user devices such as web browsers send request messages for streams, where the request messages specify a variable number of data packets of the streams for the server to send. Each of the data packets include one or more frames, or frame data, of the streams. The streaming server then “pushes” the requested number of data packets of the streams, and the client application can adjust the number of data packets for the server to send in subsequent stream request messages to optimize the bandwidth of the communications channel.