proxy for xmlrpc calls

[Xavier]

I'm attempting to write a proxy for xmlrpc calls.

I'm starting from this code;

class MagicObject:

     def __call__(self,*args,**kwargs):
         return MagicObject.__dict__['_stop'](self,self.n,*args,**kwargs)

     def __getattr__(self,name):
         if name in ('__str__','__repr__'): return lambda:'instance of 
%s at %s' % (str(self.__class__),id(self))
         if not self.__dict__.has_key('n'):self.n=[]
         self.n.append(name)
         return self

     def _stop(self,n,*args,**kwargs):
         self.n=[]
         return self.default(n,*args,**kwargs)

     def default(self,n,*args,**kwargs):
         return 'stop',n,args,kwargs

################################################################

 >>c=MagicObject()
 >>x=c.beubeb.zzzzz(1,2,3,a='bbb')
 >>print x
('stop', ['beubeb', 'zzzzz'], (1, 2, 3), {'a': 'bbb'})

I did not write this, the source is here:
http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/435757

I want to expand this to do something like;

 >>a=MagicObject()
 >>x = a.b.c.d

Then on the last __getattr__ send a call over xmlrpc.
How do I determine when the last __getattr__ will be?

With method calls you know to send a call on xmlrpc when you hit 
__call__.  But if you're dealing with a getting nested attributes
what do you do?  How do I know when there are no more attributes?
I thought about getting the source code from
the correct frame, but I don't think thats a good solution.

Any ideas?
Thanks in advance.