Question about generics

In my previous post on Generics, I received a comment from Eddie, who
was asking why developers would want to use the following constraint,
in the example below, when he would pass the parameters as IComparable
interface types.

public static bool IsLessThan(T one, T two) where T:IComparable
return one.CompareTo(two) < 0;

public static bool IsLessThan(IComparable one, IComparable two)
return one.CompareTo(two) < 0;

The previous comment touches on the generics vs base class debate, of which there is pros and cons to both.

Eddie raises an interesting question, and I am giving it some thought.
A few reasons for not using IComparable as the parameter type come to

If you wanted to compare value types, casting to IComparable would
convert the value types to a references by boxing, which is what
generics are trying to avoid.

A reference is 4 bytes in length, so, when comparing two short
values,  Eddie’s example would pass at least 8 bytes across the
stack (+ overhead). Using a generic with Unit16 type would only use of
4 bytes for the two parameters.   

Eddie’s example would require up casting the interface reference type to
known type if the method did more than just compare two values with
CompareTo. The whole point of generics is that the developer of a
generic should not have to know anything about the intended type used
in the generic context. Otherwise, we might as well use cast from
System.Object like ArrayList does.