The Java 2 security model offers fined grained access control which can
impact runtime performance. Currently the only real optimization made
is in the removal of null and duplicate adjacent protection
domains. We currently do not have sufficient metrics to know whether
this optimization is worth the effort or whether there are
heuristics which we can use to add feedback to the optimization
algorithm.

Other forms of optimizations may also be possible. For example, it may
be advantageous to construct the contents of PermissionCollections in
the order of most frequently used Permission. Additionally, the
backing store to most if not all PermissionCollections is typically a
Hashtable. When instantiating the Hashtable we rely on default load
and capacity settings which may be suboptimal. Using a more appropriate collection class could improve performance as well as
supply better synchronization semantics.