The best idea-0 approach would be to record Ambisonics with a tetrahedral array, and ship B format or a derivative to the client. In theory that could be proc

The best idea-0 approach would be to record Ambisonics with a tetrahedral array, and ship B format or a derivative to the client. In theory that could be processed just-in-time to produce an appropriate stereo image based on head tracking. http://en.wikipedia.org/wiki/Ambisonics http://en.wikipedia.org/wiki/Ambisonics In practice the computational overhead would likely be on par with graphics processing and you might want an APU :)