News of this is getting out, so I may as well make it official. I'm working on making the elisp engine (mostly eval.c) faster, using ideas supplied by Ben, and a few of my own. If anyone else is making changes to this part of the code, talk to me first so we don't collide. By the way, I don't think that the lisp engine itself is that much of a bottleneck in XEmacs, so we should not be looking at a replacement lisp engine specifically for a blazing fast XEmacs. The best things to look at from a performance point of view, judging from my profiling runs, are - redisplay - redisplay some more - garbage collection - extent handling - regular expression handling - mule coding system character conversion Personally, I've always been interested in the language engine itself. I've been working my way up to hacking Feval() itself for 3 years now, and I just got the courage to modify that function today. Even if the core lisp is not THE performance bottleneck, those functions ought to be an example of coding excellence, and so I think they're worth a little more effort. Martin