This is very similar to the "lazy strings" idea.

unicode_internal codec uses Py_UNICODE instead of the real internal
representation (PEP 393: Py_UCS1, Py_UCS2 or Py_UCS4) for backward
compatibility.

when called with a timeout.  Patch by Arnaud Ysmal.

when called with a timeout.  Patch by Arnaud Ysmal.

getcwdb() => getcwd()

 * unicode_decode_call_errorhandler() now supports the PyUnicode_WCHAR_KIND
   kind
 * unicode_decode_call_errorhandler() calls copy_characters() instead of
   PyUnicode_CopyCharacters()

Issue #13237: Fix formatting error - the legacy shell commands weren't meant to be under the Notes heading

Issue #13237: Forward port subprocess module updates and explicitly document UTF-8 encoding assumption when universal_newlines=True

http://mail.python.org/pipermail/python-dev/2011-November/114347.html