ttims@watdcsu.waterloo.edu (Tracy Tims) (02/28/88)
In article <2844@slovax.UUCP> dale@slovax.UUCP (Dale Thomas) writes: >> Lack of a MMU us going to kill the ST for any _real_ Operating Systems! > >I have to disagree a little about the lack of an MMU and a _real_ operating >system. There are a lot of things you don't need on a computer. You don't need a large linear address space, undifferentiated registers, a 32 bit ALU, or a stack. However, each of these is valuable out of proportion to its cost. With them we can easily run large programs with large data structures, we can have compilers generate nice expression computation code, and we can stack allocate frames for subroutine instantiation. So it is with an MMU: it allows flexibility in memory allocation; it protects the operating system and individual process memory spaces. If Atari's had MMUs they would be great, instead of pathetic. [Insert standard TOS disgust.] These features are essential for building reliable and powerful operating systems. I would not have much faith in a UNIX lookalike that didn't have hardware memory protection, and I wouldn't use one in an application where reliability and resistance to attack were a concern. It is my belief, that in these days of advanced CS (cough), no software or hardware designer should take seriously (in a technical sense, ignoring crass economic issues) a computer that doesn't have an MMU. It's like the old saw about the dancing bear: it's not how well operating systems work with no MMUs, it's that they do it at all. Tracy "message passing forever" Tims (ttims@watdcsu.waterloo.edu, mollusc@csri.toronto.edu, TTIMS@WATDCS.BITNET) PS. I thought the debate about multiprocessing was pretty funny, too. It was like listening to a bunch of virgins tell me I don't really need sex. PPS. I'd love to have Coherent on my ST.