tmb@tardis.UUCP (Thomas M. Breuel) (01/08/86)
||what about all the relocation fancy footwork that unix does ||to handle the per process data area |< Lots more "reasons" that UNIX needs an MMU > ||simon kenyon |Your right, UNIX DOES need an MMU. I'll call Microsoft and H-P immediately |and tell them to stop selling Xenix and HPUX because they can't possibly |work. Boy are they going to be pissed. Oh well, they should have asked |if they needed an MMU before they went ahead and ported UNIX to hardware |that doesn't contain an MMU. Is it possible that you stop this fruitless discussion about whether 'UN*X' (whatever that means) needs an MMU (whatever that means)? Yes, there are operating systems out there that you can call UN*X and that get along without an MMU. Yes, none of the Berkeley versions of UN*X will run comfortably without an MMU (although, as Berkeley shows again and again, blind hacking can achieve anything, even if it crashes all the time and is incredibly slow...). One concluding remark: the 8086 (without an MMU) is probably more suited to running UN*X than the 68000 without an MMU. The reason for this is that the segmentation of its address space gives you some kind of rudimentary memory management. You can, in principle, do the same thing on the 68000 using indexed addressing and PC relative addressing, but you sacrifice a lot of the power of the 68000 that way. Now, you all who claim that you can make a UN*X run on the ST would do everybody a great service if you carried out what you claim is possible. Failing that, I am personally eagerly awaiting OS9. Thomas.