Hrvoje> Martin Buchholz <martin(a)xemacs.org&gt; writes: Hrvoje> No no, this is wrong. First, funcalls are very slow in Emacs, so Hrvoje> things would slow down for a large number of argument. Second, macros The idea would be that only the 2-arg operator actually has its own bytecode assigned. Byte-compiled code would always use that. Interpreted code would use a subr, as implemented and currently commented out in data.c. Hrvoje> lose with things such as `apply'; in CL, the following code returns Hrvoje> non-nil if LIST is monotonous: Hrvoje> (apply '< LIST) I agree this has to continue to work. Hrvoje> If we don't implement the CL functionality fully, it's better not to Hrvoje> bother with it at all... Doing it completely correctly is a little tricky, but worth doing. Maybe the next time I want to learn more about the byte compiler... Martin