Direct numerical simulation methods have played a growing role, albeit a background one, in musical sound synthesis for some time now. Yet, they appear to be the most straightforward and general approach to physical modeling. For such general techniques applied to complex nonlinear systems, the problem of numerical stability lies on the horizon. This article has attempted to outline a more modern time-domain technique for the analysis and construction of physical modeling sound synthesis algorithms, which addresses this issue, as well as some of the more delicate questions of boundary termination. Neither of these is dealt with adequately using frequency domain methods. In essence, the closer attention one pays to the underlying dynamics (particularly the energetic properties), the more robust a method may be constructed