After thinking about it a bit more I am OK with Vinay's proposal.  Attached is an updated patch.

Also, I also noticed that the 'struct' module has the same problem:

>>> big_int = int(sys.float_info.max) * 2
>>> struct.pack('d', big_int)
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
struct.error: required argument is not a float

but the 'array' module does the right thing:

>>> big_int = int(sys.float_info.max) * 2
[68068 refs]
>>> array.array('d', [big_int])
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
OverflowError: long int too large to convert to float
[68068 refs]
>>> array.array('d', [""])
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
TypeError: a float is required

Mark, do you have any opinions on the error handling here and in the struct module?