wbe@BBN-VAX.ARPA (11/17/85)
From: "Winston B. Edmond" <wbe@BBN-VAX.ARPA> There seems to be this persistent belief that memory management is needed to get memory protection. It's not true. Memory management allows one to have segmentation and/or paging, and has some kind of memory mapping arrangement. Manufacturers of inexpensive, high performance systems look at these and say "Address translation will slow down the system and cost more money. Let's not do it." They may even be right, in the sense that their intended market may not need it. Memory protection can be as simple as base-bounds registers, which specify the lowest and highest allowed addresses. It does not require address translation, and does not slow down the machine. What it does do is detect out-of-bounds references and trap to the operating system. Memory protection is enough to make debugging in multi-processing systems fairly safe. Unfortunately, in my opinion, most manufacturers seem to decide that it's either full memory management or nothing. ~~flame warning~~ The view that consumers won't care because they buy working products rather than develop them, and that there aren't enough developers to make the additional cost worthwhile, overlooks something. A system that is easier for developers will make software products appear sooner. This helps sell more computers. If debugging is easier, the products will be more thoroughly tested and reliable, which makes the developers look good, which makes them happier they chose to write software for that machine. __end flame__ Are there any plans for an Amiga II that supports memory protection or memory management? -WBE
anton@ucbvax.BERKELEY.EDU (Jeff Anton) (11/20/85)
Hear! Hear! If the Amiga had memory bounds checks on processes in hardware (and low level documentation) I'll have burnt $$$ on it by now. I've been waiting for a SUN to appear in my bedroom for too long. Maybe I'm behind the times, but I want a system that will do what I command it to do. And I've the talent to command at all levels of a system. I want to be given a chance to add my own hardware and system software to support it. Having written an editor, lisp interpreter, modem communications package, etc. for my CP/M system I want to keep my software. Much of system software for small systems is convoluted or lazy. I may want to add kernel level networking or something big like that. I'm saddened to find that so many people who are getting their first tastes of programming are satisfied by blistor packed, sanatized, and user handholding systems. Give me a good compiler, a bus extender, and a logic probe instead. These small systems are approching increadable power; but, like video games, the power is always special cased or just raw. I suspect that today's Amiga if it had 1 Meg of memory + hard disk + memory protection, would surpass the performance of systems MUCH more expensive. It's inevitable I guess when the vulture capitalists target their work on one type of buyer. I want to sell the following stickers: Mac: When I grow up, I want to be a SUN Amiga: When I grow up, I want to be an IRIS ST: No respect PC: Trust me, I'm here to help you. INTEL: Which way did they go? Which way did they go? ATT: It has to be a standard, we built it. System V: If it weren't a standard, it would be sub-standard. It looks like I'll be getting a SUN in maybe a year. Or maybe I'll wait for the Cray on a chip Bill Joy predicts will exist in 1987, is in a machine. -- C knows no bounds. Jeff Anton U.C.Berkeley Ingres Group ucbvax!anton anton@BERKELEY.EDU