The first computer-aided maintenance software came from DEC in the 1980s to configure VAX computers. The software was built using the techniques of artificial intelligence expert systems, because the problem of configuring a VAX required expert knowledge. During the research, the software was called R1. As one of the first expert systems to be pressed into commercial service, renamed XCON, it created high expectations, which did not materialize, as DEC lost commercial pre-eminence.

But, fundamentally, XCON was a rule-based configuration database. The expert system used forward chaining rules.

Reference: Virginia E. Barker and Dennis E. O'Connor. Expert systems for configuration at Digital: XCON and beyond. Communications of the ACM, 32(3):298--318, March 1989.

Help Desk software

Help Desk software frequently captures symptoms of a bug and relates them to fixes, in a fix database. One of the problems with this approach, is that the understanding of the problem is embodied in a non-human way, so that solutions are not unified.

Strategies for finding fixes

  1. The bubble-up strategy simply records pairs of symptoms and fixes. The most frequent set of pairs is then presented as a tentative solution, which is then attempted. If the fix works, that fact is further recorded, along with the configuration of the presenting system, into a solutions database.
  2. Oddly enough, shutting down and booting up again manages to 'fix' (or at least 'mask' a bug) in many computer-based systems; thus reboot is the remedy for distressingly many symptoms in a 'fix database'.
  3. Currently, most expertise in finding fixes lies in human domain experts, who simply sit at a replica of the computer-based system, and who then 'talk through' the problem with the client to duplicate the problem, and then relate the fix.