> $ be new 'utf8 string'
> Traceback (most recent call last):
>   ...
> UnicodeDecodeError: 'ascii' codec can't decode byte 0xd0 in position 95: ordinal not in range(128)

(bug reported against cjb@laptop.org-20091006145647-kqkmoh481tl5hvt4)

This was fixed with revision
  wking@drexel.edu-20091117145118-jltbju9thsn5xvkv
in my branch on Nov. 17, 2009.

> I think it is more correct to use UTF-8 everywhere or use
> locale.getdefaultlocale() instead sys.getdefaultencoding().

We try to use unicode strings internally, it's input/output that's
difficult.  This particular bug turned out to be related to our
mapfile storage handling.  Take a look at the be.unicode-hg branch
leading up to revision
  wking@drexel.edu-20091117145118-jltbju9thsn5xvkv
for details.