If memory does server, I do recall the lisp hackers at MIT having native lisp machines in the wayback; and they were fairly on the kludgy side. As for "quasi- realtime", that is totaly up to what ever scheduling mechinism that you implimented presumeably on a lower ring level, and doesn't realy have any bareing on how you wrote your script. As for real time (hard or soft) scheduling with trinux, and linux in gerneral, is not obtainable without hevy modification (ie the RT patches). As for having it kernelised, which is what I think you ment, for the love of allah, why? Just becuse you could prossiably kludge micoperl's binary image objects into the kernel image doesn't by any means you should. As Cobato taught us, complex system design is bound to fail ultimately. I'll spare you the spiel about value of seperating the method form the polocy.