It is often convenient to be able to save and subsequently restore the state of a computation. It is also, in general, rather expensive. We present a technique whereby states may be saved and restored extremely fast, and yet the information stored in a given state may be accessed and updated reasonably quickly. This technique depends on an indexing scheme which can be used to see whether the last value assigned to a variable is still valid, or whether it must be retrieved from some previous context (and if so, which one).
This paper is not available online