> Subclassing
> ----------------
> One issues related to subclass structure is whether to allow 
> a class to be defined as a subclass of a class that is not in 
> its open access scope.  I would like to prohibit this, 
> although I can imagine it leading to some fragmentation of 
> the package structure.  Without this prohibition, an active 
> superclass pointer can point into the internals of a segment 
> that is not resident.  We have dealt with this in the present 
> ImageSegment work by forcing all classes in a segment to be 
> roots (ie to be converted to root stubs when not resident), 
> but it would be nice to have an architecture that did not 
> require this.

A related issue is supporting "base class changes".  I know its 
controversial in some quarters, but it certainly is common for 
packages to add a method here and there to Object,UndefinedObject,
String, etc.  This also has implications in the browsers.

-david