"Tolerating inconsistency" revisited

Abstract

We're surrounded by inconsistency: in our requirements, in the data that our software processes, and in those software systems themselves, Yet our formal systems can't handle such inconsistency. Most of them lose the ability to form any valid conclusions or analyses in the presence of even a single inconsistency. This forces our programs to operate in terms of an idealized model rather than the real world with the attendant requirement to either maintain a mapping between the two or force human operators to resolve the inconsistencies before the data is processed by the idealized system. My "Tolerating Inconsistency paper introduced a simple way to scope formal constraint systems so that they applied only to the consistent data. Data inconsistent with these rules could then be represented and processed by giving them special marks to place them outside the rules' scope. My talk will review the influence this idea had on the field and my subsequent work.