Craig:

Also, perhaps we should follow Consel and some others (and some of our own
earlier thoughts) and allow more declarative specification of which fields of 
a struct are static and which globals are static.  We'd allow procedural 
overriding of these defaults (*dynamic and *static operators; * just takes the
default interpretation for the target memory).  I guess we might think about
having static or dynamic annotations become part of the type of a variable/
field, or a storage class, or something like that.  "*static p" then would
be sugar for "*((static foo*) p)", i.e., a cast to static.  static and dynamic
fit in the same place that const does now.  Declarative specification where
it works, and procedural overrides for full flexibility.