Abstract
The halt scheme for quantum Turing machines, originally proposed by Deutsch, is reformulated precisely and is proved to work without spoiling the computation. The ``conflict'' pointed out recently by Myers in the definition of a universal quantum computer is shown to be only apparent. In the context of quantum nondemolition (QND) measurement, it is also shown that the output observable, an observable representing the output of the computation, is a QND observable and that the halt scheme is equivalent to the QND monitoring of the output observable.