Hexadecimals can be formatted to lower and uppercase:

>>> '{0:x}'.format(123)
'7b'
>>> '{0:X}'.format(123)
'7B'

I would like the same thing for strings:

>>> '{0.lastname:u} {0.firstname}'.format(user)
'DOE John'

I first thought using "S" for uppercase, but "s" is not available for lowercase. So I thought about "u" and "l".

The alternative is to write:

>>> '{0} {1}'.format(user.lastname.upper(), user.firstname)
'DOE John'

But I find it less compact and elegant.

The format support is written specifically so that it is extensible.  You can write your own string subclass that extends the formatting mini-language with whatever features you find useful.  There are too many variations on what might be useful with regards to case transformations for this to be a sensible addition to the language itself.  Much better that an application create use-case tailored facilities.

(Note, by the way, that new features can only go in to 3.3 at this point.)

I agree with David.

Here's an example of using such a subclass. It extends the format string for strings to begin with an optional 'u' or 'l':
-----------------------
class U(str):
    def __format__(self, fmt):
        if fmt[0] == 'u':
            s = self.upper()
            fmt = fmt[1:]
        elif fmt[0] == 'l':
            s = self.lower()
            fmt = fmt[1:]
        else:
            s = str(self)
        return s.__format__(fmt)

name = 'Hervé Cauwelier'

print('{0:u*^20} {0:l*^20} {0:*^20}'.format(U(name)))
-----------------------

It produces:
**HERVÉ CAUWELIER*** **hervé cauwelier*** **Hervé Cauwelier***

Thanks for the example. The Python 2.7 documentation about the mini-language doesn't clearly state that it is extensible and how. we see examples of formatting but not of extending.

Your example would be welcome in the documentation.