As stated at the beginning of Section 4, the user would ideally want to be asked a few questions about his or her problem, and in return get summary of the right algorithm to use, a pointer to corresponding software, and performance and accuracy predictions. In addition to the conventional book format in which such information could be presented, we plan to explore the use of a hypertext interface to let the user browse through the information, and traverse the underlying decision tree. Both ``black boxes'' as in LAPACK and ``templates'' will be possible recommendations at the bottom of the decision tree. Black boxes will be briefly described, but not in sufficient detail to reproduce a good implementation (which is why they are black boxes!). Templates will be made available in pseudocode, Matlab, and Fortran, as they were in the prior Templates book. If there is strong feeling that C, C++, or Fortran 90 should also be used, we would like to hear it (but note that we do not have the resources to recode LAPACK in Fortran-90 or C++, although wrappers are possible). We hope to assemble a set of test cases, and evaluate the algorithms we suggest on these test cases. The test cases should demonstrate both typical and extreme behaviors of the algorithms.