“…Once again, several examples of computations satisfying this condition have been published. These examples, are less well known as they concern nonstandard, yet realistic, paradigms, including, for example, problems where (i) All the data are not available at the outset of the computation, but instead arrive over time; the computation is considered complete when all the data arrived so far have been handled regardless of whether more data arrive later [6,7,24,25], (ii) The values of the data change as the algorithm proceeds; the computation is considered complete when all the corrections arrived so far have been handled regardless of whether more corrections arrive later [8,24,25], (iii) A computation is involved that is not efficiently invertible: With a sufficient degree of parallelism, the inverse computation is not required; with an insufficient number of processors, the inverse computation becomes necessary [2]-…”