The algorithms in MINTOOLKIT were chosen based upon many sources of information, two of which are Nocedal (1992), for continuous, convex problems, and Mittelmann for global minimization. The goal of MINTOOLKIT is to be able to solve well-posed problems quickly and robustly, using the smallest set of algorithms possible. ''Well-posed'' is an important adjective here - MINTOOLKIT is not meant to be able to solve poorly conditioned problems or problems that can easily fall into numeric precision traps. Please pay attention to how your data is scaled, try to ''bullet-proof'' your objective function to avoid divisions by zero, etc. Nevertheless, if you find bugs, or have suggestions or comments, please contact the author.