In all of the tools we have described so far, users either manipulate an artifact privately or in a joint session. CAIS (Collaborative Asynchronous Inspection System) [] illustrates the benefits of intermixing of these two modes of collaboration. Traditional inspections assume that the complete inspection task will be done in meetings. CAIS changes this paradigm to allow part of this task to be done in asynchronously, thereby reducing the overhead of scheduling meetings. The application supports discussion threads, voting on faults, and fault summaries. At the end of a user's individual asynchronous session, other users are automatically sent mail regarding the comments and votes entered by the user. If any user votes to hold a meeting, then that fault is set aside for the next synchronous meeting. Moreover, during a meeting, the application allows a comment to be broadcast incrementally, on every keystroke, or in batch when it is committed.
Like CES, CAIS distributes the shared information among distributed autonomous objects. In particular, it creates a separate distributed object for each annotation, the document being browsed, the history log, the hierarchical discussion threads, the collection of faults, and the inspection summary. It is built on top of the Suite infrastructure (which we shall study later) and uses its distributed object primitives and flexible coupling facilities for implementing its features.
Experience with CAIS shows that most faults can be identified asynchronously, and meeting are held only for the few controversial cases that occur. It also shows that the inspection process takes longer when asynchronous inspection is allowed but finds more faults. The duration of the inspection process was measured by determining the periods during which a user was active. It is not clear what fraction of these periods users actually performed useful work.