* Review PDD 19 PIR for namespaces behavior
 * Review PDD 21 Namespaces

= NameSpace PMC =

 * NameSpace has too many named-storage responsibilities: subs/methods by name, nested namespaces by name, and global variables by name. These responsibilities should be broken up and accessed through separate interfaces. 
    * Remove FPA_is_ns_ext flag nonsense
    * Stop using get/set_pointer_*. 
 * NameSpace should cache it's own fully-qualified name in a standard format (RSA, FSA, etc)
 * Decouple NameSpace and Class. One should not imply or require the existence of the other
    * Stop marshaling Sub/methods from NameSpace to Class. Or, find a much less expensive way to do it. For instance, if we have a specialized "sub cache" object, we can simply move that sub cache from the NameSpace to the Class.
 * get_namespace_pmc should be renamed to indicate that it operates on subs. It should probably not poke directly into the sub guts
 * del_* methods should be renamed to "delete_", for clarity

= src/namespace.c =