Constraints from STRIPS - Preliminary peport

We re-visit the problem of converting action specifications into system constraints by re-examining it in the very simple setting of STRIPS. This has the merit of making many thorny issues relatively transparent. The paper is in the form of an extended summary of ongoing work and many of the results are merely outlined. But sufficient details are included to indicate where we will be taking things further, and to encourage others to pursue the same objectives. These objectives are in some sense a kind of reverse-engineering, as the database community is evidently familiar with the idea of starting with constraints and then deriving action specifications from them. However in AI reactive systems, action specifications are often the primary entity, so techniques to unearth the implicit constraints can facilitate better designs.