What is the difference between Python's random.gauss and normalvariate functions?

I need to implement a gauss function and a normalvariate function. I have a good source of entropy, and I want to use that to mimic the behaviour of the built-in random module.

I have implemented gauss by doing a Box-Muller transform. How can I map my 64 bits of randomness to a normalvariate distribution? Or is Box-Muller close enough for most purposes that I can just alias it?