On 8/12/2012 1:36 AM, Johannes Pfau wrote: > Probably we should distinguish between finish functions which destroy > the internal state and peek functions which do not modify the internal > state. I worry about too many parts to an interface. A peek function needs a really strong use case to justify it.