The BaseManager.from_address method is documented at:

http://docs.python.org/dev/library/multiprocessing.html#multiprocessing.ma
nagers.BaseManager.from_address

but it looks as though this method doesn't actually exist.  In contrast, 
the BaseManager.connect method appears to be necessary for making remote 
connections, but is not documented.

The documentation should be amended as follows:

Running the following commands creates a server for a single shared
queue which remote clients can access:

>>> from multiprocessing.managers import BaseManager
>>> import Queue
>>> queue = Queue.Queue()
>>> class QueueManager(BaseManager): pass
...
>>> QueueManager.register('getQueue', callable=lambda:queue)
>>> m = QueueManager(address=('', 50000), authkey='abracadabra')
>>> s = m.get_server()
>>> s.serve_forever()

One client can access the server as follows:

>>> from multiprocessing.managers import BaseManager
>>> class QueueManager(BaseManager): pass
...
>>> QueueManager.register('getQueue')
>>> m = QueueManager(address=('localhost', 50000), authkey='abracadabra')
>>> m.connect()
>>> q = m.getQueue()
>>> q.put('hello')

Another client can also use it:

>>> from multiprocessing.managers import BaseManager
>>> class QueueManager(BaseManager): pass
...
>>> QueueManager.register('getQueue')
>>> m = QueueManager(address=('localhost', 50000), authkey='abracadabra')
>>> m.connect()
>>> q = m.getQueue()
>>> q.get()

Also, it would be helpful to elaborate a bit more on:

major:
 * how to implement a queue that is shared both locally and remotely
(i.e. n local processes access the queue as well as m remote processes)

minor:
 * blocking (assumption: q.get() blocks on socket.recv())
 * timeout (assumption: the socket does not time out, e.g. q.get()
blocks endlessly on socket.recv())

I propose we add the following to that section as well.

If you need to provide access to a queue from both local and remote
processes, use `multiprocessing.Queue` in the server:

>>> from multiprocessing import Process, Queue
>>> from multiprocessing.managers import BaseManager
>>> class Worker(Process):
...     def __init__(self, q):
...         self.q = q
...         super(Worker, self).__init__()
...     def run(self):
...         self.q.put('local hello')
... 
>>> q = Queue()
>>> w = Worker(q)
>>> w.start()
>>> class QueueManager(BaseManager): pass
... 
>>> QueueManager.register('getQueue', callable=lambda: q)
>>> m = QueueManager(address=('', 50000), authkey='abracadabra')
>>> s = m.get_server()
>>> s.serve_forever()

Use it in the client as shown above:

>>> from multiprocessing.managers import BaseManager
>>> class QueueManager(BaseManager): pass
... 
>>> QueueManager.register('getQueue')
>>> m = QueueManager(address=('localhost', 50000), authkey='abracadabra')
>>> m.connect()
>>> q = m.getQueue()
>>> q.get()
'local hello'
>>> q.put('remote hello')

Added the mp.managers shared queue example, fixed the docs in r67419 on 
trunk. merged to py3k and 2.6.1 maint

I would like to open this ticket back up.

Python 2.6.2 docs still reference unimplemented 'from_address' method.

http://docs.python.org/library/multiprocessing.html#multiprocessing.managers.BaseManager.from_address

BaseManager.__reduce__ method also calls unimplemented 'from_address'
method. This looks like a bug since there is no docs or comments that
indicate it is as an abstract method.

New changeset c2910971eb86 by Richard Oudkerk in branch 'default':
Issue #3518: Remove references to non-existent BaseManager.from_address()
http://hg.python.org/cpython/rev/c2910971eb86