Really fast sort on semi sorted data

Good question (my 5), but I think the answer would depend on the definition of semi-sorted. You see, different algorithms works with different speeds on different data sets. When the data is random (still it would need the description of randomness, such as average size, total number...) you can do comparison. As soon as you say, "semi-sorted", there are different ways of getting this random semi-sorted state -- and comparison of algorithms can be different for different ways. Isn't it logical? --SA