Proposed Model for Presentation without VSYNC

I Believe the Oculus SDK could really benefit from an additional configuration flag that denotes the SDK is being used with no VSYNC. If the Oculus SDK is then told explicitly that DirectX / OpenGL is not using VSYNC I believe it will work very well if EndFrame() takes a timestamp, adds it to the accumlated average(minus this frames wait time) and calculates a continuous average of n frames, then calls Present().
BeginFrame() then delays or waits based on this average, minus say 1ms or 2ms to compensate for error.
I think that the Model Needs to be Changed at very fundamental level to minimize Input Lag. This model presented would achieve significantly better response to have Input equal Output.
I was hoping an Oculus Developer would chime in specifically if this can be done with the SDK. I would like to know, as it could be done manually in my application, but it seems more natural that the SDK do this for you.
Thanks!
