Hrv> Totally off-topic, but if someone mentions the "30%" figure once Hrv> again, I'll SCREAM! It's a myth. Subjectively, Mule is at least 3 Hrv> times slower than non-Mule -- and Martin's benchmarks confirmed that Hrv> "30% slowdown" is an insulting euphemism. ?? 3 times slowdown is not my experience. I run a Muleified XEmacs as my standard editor. Maybe you have strong memory constraints, and increasing the size of XEmacs only slightly causes large increases in swapping. Or perhaps you regularily use an application that falls prey to some N*N algorithm that is very easy to get in Mule XEmacs (I've fixed a few of those, but I'm sure some remain). Hrv> Not that I have any problem with what you, Olivier, are saying here -- Hrv> I just want to draw everyone's attention that repeating the 30% myth Hrv> won't make it true. Sorry Hrvoje, but 30% (in 21.2) feels about right to me. Martin