Abstract

We present Persistent Turing Machines (PTMs), a new way of interpreting Turing-machine computation, one that is both interactive and persistent. A PTM repeatedly receives an input token from the environment, computes for a while, and then outputs the result. Moreover, it can \remember" its previous state (work-tape contents) upon commencing a new computation. We show that the class of PTMs is isomorphic to a very general class of eective transition systems, thereby allowing one to view PTMs as transition systems \in disguise." The persistent stream language (PSL) of a PTM is a coinductively dened set of interaction streams : innite sequences of pairs of the form (w i ; w o ), recording, for each interaction with the environment, the input token received by the PTM and the corresponding output token. We dene an innite hierarchy of successively ner equivalences for PTMs over nite interaction-stream prexes and show that the limit of this hierarchy does not coincide with PSL-equivalence. The presence of this \gap" can be attributed to the fact that the transition systems corresponding to PTM computations naturally exhibit unbounded nondeterminism. We also consider amnesic PTMs, where each new computation begins with a blank work tape, and a corresponding notion of equivalence based on amnesic stream languages (ASLs). We show that the class of ASLs is strictly contained in the class of PSLs. Amnesic stream languages are representative of the classical view of Turing-machine computation. One may consequently conclude that, in a stream-based setting, the extension of the Turing-machine model with persistence is a nontrivial one, and provides a formal foundation for reasoning about programming concepts such as objects with static elds. We additional...

Citations