Well I'm not subscribed to the python-3k list either - too much traffic
indeed. You can read and post into it with gmane for example:
http://thread.gmane.org/gmane.comp.python.python-3000.devel/11768
(there is probably an NNTP gateway too)

As for instrumenting the interpreter, this would tell us when and which
strings are allocated, but not the precise effect it has on a modern
CPU's memory subsystem (which is quite a complicated thing). Also, this
is Py3k - we can't test any real-life applications until they are ported
to it...
(someone really motivated would have to learn Intel VTune or AMD
CodeAnalyst and run such a "py3k real-life application" with it :-))

As for the explicit slicing approach, "explicit string views" have been
discussed in 2006 on the python-3k list, including a proof-of-concept
patch by Josiah Carlson - without any definitive pronouncement it seems.
See the subthread here:
http://mail.python.org/pipermail/python-3000/2006-August/003280.html

The reason I'm bringing in those previous discussions is that, in
theory, I'm all for much faster concatenation and slicing thanks to
buffer sharing etc., but realistically the previous attempts have failed
for various reasons (either technical or political). And since those
optimizations are far from simple to implement and maintain, chances are
they will not be attempted again, let alone accepted. I'd be happy to be
proven wrong :)

regards

Antoine.