The key problem here is that difficulty has two elements:

 - The question
 - The person answering it

Now lets assume you can perfectly gauge the question part, you can't account for the person answering it.  To take an extreme example, I consider myself a reasonable Java developer, I am however not a poet. So I would consider the first question much easier than the second

 - How can I create a persistent file that represents objects in a binary form?

 - What is iambic pentameter? 

A different person (in fact most people) would probably feel the opposite.

This is clearly an extreme example, but one persons easy question is another person's impossible question.  As such such a measure isn't very helpful