Jan Vroonhof <vroonhof(a)math.ethz.ch&gt; writes: It isn't based on the "FSF font model", as there is no such thing. It is based on the pre-specifier *face* model, plus a subset of wmperrys font.el package, the closest thing you get to a *font* model on either emacs. The unbundled code doen't *touch* specifiers, as they are unnecessary for what it tries to do. It uses the pre-specifier API's together with `make-frame-hook', which is sufficient. I suspect most of the current problems is caused by a port of the code to specifiers, that only got halfway finished. There is no XEmacs font model, which is the only hard thing about it. I discussed with wmperry how to make it use the full font.el font model, that's not hard, it is just a small matter of programming. On both emacsen. But you can overwrite the customize setting with X ressources, which only leaves explicit Lisp settings. That problem is orthogonal to specifiers, which is why we should talk. I believe You think about the problem the wrong way, and I think the person doing the XEmacs reimplementation did the same, and that that is the cause of the current problems.