NOTE: This transcription was contributed by Martin P.M. van der Burgt, who has devised a process for producing transcripts automatically. Although its markup is incomplete, we believe it serves a useful purpose by virtue of its searchability and its accessibility to text-reading software. It will be replaced by a fully marked-up version when time permits. —HR


Copyright Notice
The following manuscript
        EWD 675 The equivalence of bounded nondeterminacy and continuity
is held in copyright by Springer-Verlag New York.
The manuscript was published as pages 358359 of
Edsger W. Dijkstra, Selected Writings on Computing: A Personal Perspective,
Springer-Verlag, 1982. ISBN 0387906525.
Reproduced with permission from Springer-Verlag New York.
Any further reproduction is strictly prohibited.
                                                                                                                                               

The equivalence of bounded nondeterminacy and continuity.

Unbounded nondeterminacy is presented by the function “any natural number” such that

wp("x:= any natural number", 0 ≤ x) = T
wp("x:= any natural number", x ≤ k) = F for all k .
Program S is continuous —see Chapter 9 of “A Discipline of Programming”, where this property is called Property 5— means that for any infinite sequence of predicates C0 , C1 , C2 , ... such that
for r ≥ 0     Cr ≠ Cr+1     for all states
we have for all states
wp(S, (E r: r ≥ 0: cr)) = (E s: s ≥ 0: wp(S, cs))(1)
and in the same chapter I have shown that all programs that could be written in my programming language fragment —with finite (1) guarded command sets— are continuous.

It is further shown that the program “x:= any natural number” is not continuous —and, hence, cannot be written in that programming language fragment— . For the sake of completeness, we repeat the proof. Assume the program S: “x2: any natural number” to be continuous. We then have:

                T = wp(S, 0 ≤ x)
    = wp(S, (E r: r ≥ 0: 0 ≤x ≤ r))
= (E s: s ≥ 0: wp(S, 0 ≤ x ≤ s))
= (E s: s ≥ 0: F) = F
a contradiction that leads to the conclusion that “x:= any natural number” cannot be continuous, i.e. that continuity implies bounded nondeterminacy.

In the sequel of this note we shall show that the inverse holds as well, viz. that the existence of a noncontinuous program implies the inclusion of unbounded nondeterminacy. (The following argument was suggested to me by C.S.Scholten almost instantaneously when I had posed the problem.) Assume the existence of a program S and an infinite sequence of predicates C such that Cr ⇒ Cr+1, such that (1) does not hold. Because in (1) the right-hand side implies the left-hand side trivially, this means that we assume

wp(S, (E r: r ≥ 0: cr))and non (E s: s ≥ 0: wp(S, cs)) =
wp(S, (E r: r ≥ 0: cr))and (A s: s ≥ 0: non wp(S, cs))(2)
to be different from F .

Consider now the program

              S; x:= (MIN: k: Ck)
started in an initial state satisfying (2). Because the initial state satisfies wp(S, (E r: r ≥ 0: Cr)) , this program terminates and is guaranteed to establish 0 ≤ x . On the other hand, the assumption that for some K it is certain to establish x ≤ K means that S is certain to establish CK , a conclusion that is incompatible with the second term of (2). Hence its nondeterminacy is unbounded. (The fact that our program of unbounded nondeterminacy is not a total program, but only defined for initial states satisfying (2) is here not relevant: the essential thing is that (2) differs from F , i.e. that the set of states satisfying (2) is not empty.)

Here we have established the equivalence of continuity and the boundedness of nondeterminacy. In EWD673 we have established the equivalence between the boundedness of nondeterminacy and the equality between weak and strong termination. Hence the three criteria
1)     continuity or not
2)     nondeterminacy bounded or not
3)     weak and strong termination equivalent or not are three different aspects of the same dichotomy. All this is very satisfying. (The arguments are so simple that, presumable, this is already known. But it was new for me, and I like the arguments.)

Plataanstraat 5prof.dr.Edsger W.Dijkstra
5671 AL NUENENBURROUGHS Research Fellow
The Netherlands

Transcribed by Martin P.M. van der Burgt
Last revision 2015-02-05 .