(sorry for the delay in replying: I missed your reply because it was addressed only to the list not to me)

Except the problem of time machines: When interpolating around a non-delayed signal, you are interpolating around the time "now" and so you would need at least one sample from the future, whhich hasn't happened yet.

D'oh!!! Of course, stupid me!!! I was thinking about the one sample from the _past_ (which is nonsense) instead of the one from the _future_

Thanks m.