A common complaint against Scheme is that it has too many mutually incompatible dialects, and that their common subset as defined in the R5RS is too skimpy for real applications. Sticking to one Scheme dialect may be a sensible way to focus one's energies on the application being written -- but limits the number and diversity of users, which in turn limits crucial feedback. I describe a method of dialect-configuration that eliminates this disadvantage, so that an application written specifically for just one dialect can yet run on other dialects, including even a non-Scheme like Common Lisp, without imposing any tedious or arbitrary coding discipline on its author.