Here's a first attempt at Decimal.__format__; the patch is against the 
trunk, and should be forward ported as usual to 3.0, with obvious minor 
changes related to str/unicode.  It still needs some cleanup and some 
more tests, but I'm posting it now in the hope of getting some feedback.

I'm adding Raymond Hettinger and Eric Smith to the nosy list in case 
they want to comment.

Some points of interest:

 * I decided to make plain 'e', 'f' and 'g' formats (without an explicit 
precision) do no rounding:  they return a string that captures the exact 
value of the Decimal instance (though it can lose information about 
significant zeros).  So format(x, 'e') is basically a way to say 'give 
me str(x), but always include an exponent', and format(x, 'f') gives a 
way to print the value without ever including an exponent.  format(x, 
'g') is identical to str(x), except that it always uses the character 
'e' for an exponent (instead of using 'e' or 'E' depending on the 
context).

There are other possible options here (have a default precision;  use 
the precision from the context), but this one seemed to make most sense.  
I'd appreciate opinions.

 * the integer format specifiers are not supported.  After noticing that
the integer format specifiers aren't supported for floats either, I 
think this is the right thing to do.

 * the 'n' format specifier is not supported either;  it's supposed to 
use the current locale.  I can't find any easy way to do this---it seems 
as though the only real option is to manually fix the decimal point 
character, figure out where to place thousands separators, etc.  This 
would add quite a lot of not-really-Decimal-related code to decimal.py, 
and I'm reluctant to do that.  The right solution probably involves 
writing some support code and putting it somewhere else in the std. lib.

 * on the subject of not-really-Decimal-related code, it would be great
if the parse_format_specifier and format_align functions (near the end 
of decimal.py) could be moved somewhere else.  I'm planning to reuse 
these functions for Fraction.__format__.