Issue 73, where does reduction start?

All,

Issue 73 is about at which level in a nested policy set does delegation
reduction start?

As it is specified currently, reduction is done as a graph search where
the nodes of the graph are the policies in a policy set. In other words,
we could say that "reduction starts at the siblings of the policy being
reduced" to use the wording of issue 73.

Alternatively it has been suggested that reduction should start at the
top level in the PDP, leading to recursive invocations of the PDP.

The reduction algorithm was designed as it is currently for two reasons:

1) The current mode gives an algorithm which is guaranteed to terminate
(unless the 'access-permitted' function is invoked in the policies) and
is much easier to analyze and comprehend than the "top level PDP"
alternative.

2) There are security issues related to policy sets with nested issuers
in case of a "top level PDP" approach. In effect the issuer of a policy
set with nested