We address the problem of maintaining the integrity of large knowledge bases using a
compile-time transaction modification technique. The novelty of the approach lies in the
adaptation of ideas from Artificial Intelligence (AI) planning research. Specifically, starting
with the observation that solutions to the frame and ramification problems can be used
during database transaction design time, we propose an integrity maintenance technique that
modifies transaction specifications by incorporating into them conditions necessary of the constraints'
satisfaction. Additions to the transactions' postconditions whose effect is to maintain
the integrity constraints, are generated from a set of (determinate) transaction specifications.
Thus, the implications of constraints are realized by the transaction specifier and the effort
of having to prove transaction safety is saved, since it is guaranteed by the correctness of the
generation process. We envision the development of a tool that, given a set of determinate
transaction specifications, automatically suggests additions to the transaction postconditions
whose effect is to maintain the integrity constraints.