Title: Developer's Issues Encountered from Writing The VGAM package Author: Thomas Yee, University of Auckland Abstract: VGAM is a large R package that fits many models and distributions, usually by maximum likelihood. The key ideas are Fisher scoring, iteratively reweighted least squares and vector smoothing. The package has been under development now for more than 10 years, and I hope to discuss some issues/topics that I have encountered along the way as a developer. These include smart prediction, Fortran 77/95 and C, S4 classes (especially inheritance), cross builds, publications, and S-PLUS.