An important behavioural property for sets of active database rules is that
of termination.  In current commercial database systems, termination is
guaranteed by imposing a fixed upper limit on the number of recursive rule
firings that may occur.  This can have undesirable effects
such as prematurely halting correct executions.  We describe a new approach
based on a dynamic upper limit to the number of rule firings.  This limit
reflects knowledge about past rule behaviour on the database and provides a
more accurate measure for when the DBMS should terminate rule execution.
The approach incurs little cost and can easily be integrated with current
techniques for static analysis of active rules.