Assigning a polarity to zero is an accepted practice that allows otherwise insolvable problems to become tractable.
Jenn has guided the readers of this thread in that direction and she is correct in what she has said so far. She is also correct in that this is a concept that is much easier to handle when Multi-variable Calculus is part of your routine mathematical tool box.
“BALDERDASH!!!” you say?
First we can start with how it is represented.
http://en.wikipedia.org/wiki/−0_(number)#Representations
I think it is worth noting that the concept of negative zero is useful enough that is is defined in 1+7 bit sign-and-magnitude representation, simple binary, one’s complement, and the IEEE 754 standard.  Jenn is not alone in her interest (fascination?) with signed zero.
“I don’t care that you can write negative zero, what is it’s value?”
http://en.wikipedia.org/wiki/−0_(number)#Properties_and_handling
As the link describes, there are various means of comparing (-0), (0) and (+0) and getting a value ranking and different answers that are useful.
“All very interesting but what is it good for?”
http://en.wikipedia.org/wiki/−0_(number)#Scientific_uses
This is a fairly simple case where the idea of negative zero is used to describe days where the temperature is below 0 but not -1.Â
More complex examples are available but to be bluntly truthful about it, with out that multi-variable calculus background that I mentioned and an abiding interest in the topic, your eyes will glaze over.
“But I really want to know!!!”
Good for you.
Lets put together a very simple control system. We have a submarine with a compass and a rudder. We tell it to steer a course using the rudder. If the submarine points to the left of the ordered course we have the control system provide right rudder to return it toward the ordered course. We do the opposite if the submarine points to the right of ordered course. This gets expressed mathematically as “Rudder Angle = (Ordered course - Actual course)” where positive values for Rudder Angle correspond to right rudder. This works great if the ordered course is 180 and the actual course is 175. We get 5 degrees of right rudder. As long as the control loop is faster than the response time of the vehicle this will bring the vehicle back to the ordered course.
Now lets put some stress on our control system. Ordered course is 358. Actual course is 002. our rudder angle is now = +356 but this value takes us the long way around to the desired course instead of a negative (left) rudder taking us back to the ordered course the short way. The end result is your control system occasionally has the vehicle do right hand cirlces if it gets knocked off course too far. In addition, the controller attempted to drive the rudder to an angle that it could not reach. This is not desirable.
In order to avoid these kinds of issues we start to do things like take derivates and integrands of actual and ordered courses. From this we develop orders for the rudder.
http://en.wikipedia.org/wiki/Derivative
http://en.wikipedia.org/wiki/Integral
Sadly, as nice as these tools are for smoothing the wild swings of the rudder, we know have to deal with mathematical functions that will give us zeros at awkward times. Even more confusing is we also have to take into account the performance of the vehicle at different speeds, environments and configurations.
Remember the idea that the control loop needs to be faster than the vehicle response? There is a lot more mathematics behind that statement than I care to type. The end result is asymptotic behaviours. This means we need to know what the “sign” is of zero as we approach that asymptotic point in the control system.
Confused yet?
If not, consider a career in engineering. If you are confused, don’t feel bad. I probably totally goofed on the explanation somewhere.