http://docs.python.org/py3k/library/re.html
"Note that patterns which start with positive lookbehind assertions will never match at the beginning of the string being searched; you will most likely want to use the search() function rather than the match() function:"

>>> re.match(r'(?<=^)abc', 'abc').group()
'abc'
>>> re.match(r'(?<=\b)abc', 'abc').group()
'abc'
>>> re.match(r'(?<=\A)abc', 'abc').group()
'abc'
>>> re.match(r'(?<=\A)abc', 'abc', re.DEBUG).group()
assert -1 
  at at_beginning_string 
literal 97 
literal 98 
literal 99 
'abc'
>>>


in some cases match() can match

New changeset 7c262962b681 by Ezio Melotti in branch '2.7':
#14461: fix wording.
http://hg.python.org/cpython/rev/7c262962b681

New changeset 7f35da912739 by Ezio Melotti in branch '3.2':
#14461: fix wording.
http://hg.python.org/cpython/rev/7f35da912739

New changeset d68b4885fc0f by Ezio Melotti in branch 'default':
#14461: merge with 3.2.
http://hg.python.org/cpython/rev/d68b4885fc0f

Technically you are correct, however using zero-width classes inside a lookbehind doesn't make much sense, because the result would be equivalent even without lookbehind.
I replaced 'never' with 'not', because usually it will not match, except in these corner cases that can IMHO be ignored.