homepage

http://docs.python.org/dev/3.0/reference/datamodel.html#special-lookup
(After fixing the link to http://docs.python.org/3.0 at http://www.python.org/doc/ (and likewise the http://docs.python.org/whatsnew/3.0.html link.)...
The comment that __getattribute__ is "Called unconditionally to 
implement attribute accesses for instances of the class" gave me hope 
that some combination of "meta" "super" and "sub" might let me access 
__getattribute__ for expression eval('obj + another_object') despite the 
special notes. I realize now the truth is that
"__getattribute__ is NEVER accessible in pure python code when the code 
uses the syntax of a unary or binary operator such as a+b, ~a, len(a)."
See most of the functions in this manual section.
Also name hash, which doesn't find much explicit use but could well be 
the most used python functionality. Find a smooth way to replace my 
NEVER since code can obviously access __getattribute__ from the special 
function.
Thank you, and great work!

>>> class c:
... def __getattribute__(self,*args):
... print('getattribute chimes in')
... 
>>> c()+3
Traceback (most recent call last):
 File "<stdin>", line 1, in <module>
TypeError: unsupported operand type(s) for +: 'c' and 'int'

Isn't what you refer to covered by this paragraph and the following
example: "In addition to bypassing any instance attributes in the
interest of correctness, implicit special method lookup may also bypass
the __getattribute__() method even of the object’s metaclass:"?

Yes to msg77021. However!
I'll pin the difficulty specifically to the word "may". This cost me a
lot of time.
1) Please change the phrasing you quoted to
"... implicit special method lookup bypasses the __getattribute__()
method even of the object’s metaclass:"
2) Please insert into the glossary a definition of "implicit special
method lookup" that addresses the grammar or syntax that causes it
considering monad and dyad use.
Thank you.

I've changed "may bypass" to "generally bypasses". It doesn't happen for
all methods, but that's an implementation detail.
I've also added a glossary entry "special method" in r67579.