Alas, the techniques for insuring reliable hardware are less effective than those techniques for software and, although progress in mechanizing fault-tolerant design and measuring fault behavior has been impressive, we do not yet know how to verify highly critical systems such as HAL.

As fault tolerance technology rapidly diffuses into such continually-expanding areas as database and telecommunications systems, and is driven by applications in banking systems, medical systems, mobile computing and the ever-expanding Internet, it is clear that in the future, most computing will be done, not on isolated mechanisms at finite locations, but as a highly diffused, dynamic activity in an all-pervasive information-processing web. Fault-tolerant systems will need to be highly adaptive to changing environments. In principle, however, there seems to be nothing to prevent us from making a large computer that is fault-tolerant.