When the Gregorian calendar was established, the system of leap years was included and since then, we have been messing around with the system to see that it is exact. For instance, according to the U.S. Naval Observatory (the official time keeper for the United States. There is a good website on all of this at http://psyche.usno.navy.mil/millennium/ ). A leap second was added between 1998 and 1999 to even out the time. The reason for all of this time juggling is that to keep our arbitrary system of time accurate, continual allowances must be made for the fact that the Earth travels around the sun (making a year) but is also rotating (making a day) and these rotations don't correspond exactly (i.e. there are not exactly 365 rotations of the Earth (days) in one rotation around the sun (a year)). I would suggest that you look at the website given above as it gives a very good explanation of leap years and the historic reasons for having them. The basic answer to the question is that every fourth year is a leap year except centuries that are not divisible by 400 so 2000 is a leap year while 1900 and 2100 are not. The leap year slightly overcompensates for the actual difference between the number of days in a year and the time it takes to go around the sun so there is never a need to add a second leap day.

This is an answer to the February 2000 question; I cannot support the data with references since I remember this from having read it a while ago and I don't remember the source. Also, the information that I give may be inaccurate (I don't remember how reliable the source was), but it may be good as a second opinion to contrast to whatever other people contribute:

The rotation of Earth around the Sun takes, in fact, approximately 365.25 days, resulting in one extra day each year. However, this is not exact; there are calculations available with more significant figures that allow us to say that it's actually slightly higher than 365.242. This means having around 146097 days every 400 years, which amounts to 97 extra days every 400 years, or one extra day every four years except in three occassions, which are the centenials not divisible by 400. This can be summarized as:
Rule:365 days per year.
Exception #1:366 days per year divisible by 4.
Exception #2:365 days per centennial year.
Exception #3:366 days per centennial year divisible by 400.
Again, this doesn't amount to full exactitude in the count of days per year, but it's a very good approximation. I found all the information above in a spanish magazine of general knowledge about 7 years ago, and I found it so interesting that I memorized it. Any inexactitude can be blamed either on the publisher of the magazine or on my own memory, as you choose. Thinking about it now, I have the following comments:

+Instead of adding 2 days to the year 2000, wouldn't it be easier to include millenia in Exception #3? This would result in the same numberof extra days per 2000 years, and make much more sense to me; talking about that: anybody knows whether year 1000 had 28 or 29 days in February?

+The reason of all this hussle: the rotation of Earth around the Sun does not take an exact number of days, and we try to compensate it; if you think that this is complicated, think what would had happened if instead of ~365.242 days, it took about 365.077 days. Then, one day would have to be added every 13 years. And on top of that, still we would need to compensate every God knows how many years... Aren't we lucky that a simple number such as 4 is the best close approximation?