Well, when I first thought about this I'd say breaking it down to several n-gons would be ideal, then you would have the advantages of quads/n-gons in apps that don't handle unlimited vertexes very well. With some closer thought I'm not sure how much use it would be. From a few years ago, when n-gons first started to become common, I remember quite a few softwares having problems in two distinkt areas, one would be polys with more than 8 or 10 vertexes, the other being non-convex polys. I don't know how much of an issue this is now, though, as the apps I use now (Maya and XSI) don't hve problems with n-gons at all.