Relay-Version: version B 2.10 5/3/83; site utzoo.UUCP Path: utzoo!mnetor!uunet!husc6!mit-eddie!ll-xn!ames!sdcsvax!ucsdhub!hp-sdd!hplabs!sdcrdcf!trwrb!scgvaxd!ashtate!cy From: cy@ashtate (Cy Shuster) Newsgroups: comp.sys.ibm.pc Subject: Re: A different View of the value of OS/2 - it's better than UNIX Message-ID: <336@ashton.UUCP> Date: Sun, 27-Sep-87 20:13:44 EDT Article-I.D.: ashton.336 Posted: Sun Sep 27 20:13:44 1987 Date-Received: Tue, 29-Sep-87 06:20:29 EDT References: <494@parcvax.Xerox.COM> <961@looking.UUCP> Reply-To: cy@ashton.UUCP (Cy Shuster) Distribution: world Organization: Ashton-Tate, Torrance, CA Lines: 45 Keywords: OS/2 Summary: Different design goals Gordon Lettwin, the chief architect for OS/2 at Microsoft, characterized the difference between OS/2 and Unix (at the OS/2 Developer's Conference) as the difference between multitasking, single user and multitasking, multi-user. Unix was designed as a multiuser timesharing system, and its design goal was to maximize throughput on the shared resources: disk, memory, and printers. OS/2, in contrast, has been design to maximize *response* for the single user controlling multiple tasks. The paradigm for an OS/2 application is for the process monitoring user input to create a new thread to handle a users' request, while the main thread immediately waits for further user input. An example of how these different design goals can result in different system behavior is that under OS/2, when the system detects an overload condition (e.g., memory thrashing), it can send an alert box to the (single) user, informing him of the resource overload, and then asking him which tasks to cancel to alleviate the situation -- since all tasks were started by the same user. This would obviously not pertain to the multiuser situation! Gordon freely acknowledges that MS-DOS is barely a program loader, never mind an operating system. Almost all of the sucessful MS-DOS programs to date have been those that cleverly manage to work around the performance restrictions intrinsic to DOS, by staying resident, writing directly to video memory, and so on. This has made hackers out of the lot of us, forced to write highly hardware and software-dependent code. With OS/2 and its dynamic linked libraries, we can begin to actually use the operating system to do low-level work, and the end user can run software from different companies without worrying about RAM cram and which TSR to load last. It seems certain that OS/2 acceptance is still a ways off, however. The native 8086 emulation that Microsoft persuaded Intel to put into the 386 chip is about to spawn a host of 386 control software like DeskView 2.0, allowing multiple DOS 3 sessions on a single machine, without requiring the end user to buy all new software. The end user still doesn't care what the operating system is, if he is suffi- ciently insulated from it. The current topic of file system fragility is very germane. I, for one, favor the Lisa approach, where hitting the big red switch to power off merely generates a shutdown request to the operating system, which proceeds to neatly terminate active tasks, and save off the current desktop state before signalling a solenoid to pull the switch. I can't believe how many users do the "three-finger crash" as the normal way to exit software! --Cy Shuster UUCP: ...seismo!scgvaxd!ashtate!cy DISCLAIMER: Do you have any *idea* what I would have to go through to get this approved as an official opinion???? MINE ONLY!