The argument to my function f() must implement two different interfaces that are not related to each other by inheritance, IFoo and IBar. I know of two different ways of doing this. The first is to declare an empty interface that inherits from both:

public interface IFooBar : IFoo, IBar
{
    // nothing to see here
}

public int f(IFooBar arg)
{
    // etc.
}

This, of course, requires that the classes declare themselves as implementing IFooBar rather than IFoo and IBar separately.

The second way is to make f() generic with a constraint:

public int f<T>(T arg) where T : IFoo, IBar
{
    // etc.
}

Which of these do you prefer, and why? Are there any non-obvious advantages or disadvantages to each?