For example, in:
bool eq = (1 / double.Parse("-0.0")) == (1 / -0.0);
eq will be false.
double.Parse would have to go through some trouble to explicitly ignore the sign for zero, even though not doing that almost never results in a problem.
Since I need the raw representation, I had to write my own parsing function which special-cases negative zero and uses double.Parse for everything else.
That's not a big problem, but I'm really wondering why they made the decision to ignore the sign of zero, because it seems to me that not doing so wouldn't be a bad thing.