The following are some key points I collected from reading chapter 5 in the Software Testing book (by Paul C. Jorgensen).

One of the main uses of boundary value testing is to test for mistakes (or failures)
arising from using the wrong comparision operator. It is also a specification-based
testing (we are not worried about how the software unit is implemented) technique because we test whether the outputs are correct given valid or invalid inputs.

There are four variations of boundary value testing:
1. Normal boundary value testing (Deals only with valid inputs and is based on the single fault assumption)
2. Robust boundary value testing (Deals with both valid and invalid inputs and is based on the single fault assumption)
3. Worst-case boundary value testing (Deals only with valid inputs and is NOT based on the single fault assumption)
4. Robust worst-case boundary value testing (Deals only both valid inputs and invalid inputs and is NOT based on the single fault assumption)

The number of test cases for normal boundary value testing is 4n+1.
The number of test cases for robust boundary value testing is 6n+1.
The number of test cases for worst-case boundary value testing is 5^n.
The number of test cases for robust worst-case boundary value testing is 7^n.

From the blog Software Testing – The blog about software by MegaMind and used with permission of the author. All other rights reserved by the author.