Path: utzoo!attcan!uunet!cs.utexas.edu!yale!think!zaphod.mps.ohio-state.edu!mips!apple!agate!darkstar!saturn.ucsc.edu!golding From: golding@saturn.ucsc.edu (Richard A. Golding) Newsgroups: comp.os.misc Subject: Re: OS/2, PC's, etc... Message-ID: <2555@darkstar.ucsc.edu> Date: 13 Apr 90 18:50:10 GMT References: <9004040041.AA05123@decwrl.dec.com> <19214@boulder.Colorado.EDU> <367@jove.dec.com> <19267@boulder.Colorado.EDU> <22946@watdragon.waterloo.edu> <19338@boulder.Colorado.EDU> <5613@scolex.sco.COM> Sender: usenet@darkstar.ucsc.edu Organization: University of California, Santa Cruz Lines: 98 In article <5613@scolex.sco.COM> seanf@sco.COM (Sean Fagan) writes: > >In article <19338@boulder.Colorado.EDU> wallwey@boulder.Colorado.EDU (WALLWEY DEAN WILLIAM) writes: >>In article <22946@watdragon.waterloo.edu> tbray@watsol.waterloo.edu (Tim Bray) writes: >.... the main reason for X being so slow with 4 Mb is because of >limited memory storage, and multiple copies of the same stuff. Shared >libraries should help immensely, despite the inheirent disadvantage, >speed-wise, that SysV's shared library mechanism poses. > >>> OS/2's Presentation Manager is much faster!... >>> you can run OS/2 (witch I think is better than UNIX for the average user ... >>> With OS/2, the environment power for programs parallels that of UNIX, >>> with design tools for C that make their UNIX equivalents of Vi, CC, LINT and >>> DBX, look like they are from the dark ages!... After having worked on both the internals of PM and OS/2, and X11 and some Unices, I find it really curious that people have the impression of speed from OS/2. In most cases, *when you use comparable hardware, applications, and operating system versions*, Unix and X are at least as fast as PM/OS/2 (and usually faster). One of the faults with the PM vs. X comparison is that most people are comparing the generic MIT sample server with the much-hacked (and fairly non-portable) Microsoft PM. When you compare an X server which makes use of the facilities available under, say, SCO Unix (e.g. shared memory, shared libraries, and so on) I've found that speed is often better under X. It's worth noting that X has significantly less overhead for a great many operations; for example, one need not pay the penalty for potentially retained graphics segments as one must in PM. >>As for editors, well this is a moot point. You can have >>just about any type of editor you want under OS/2! ... > >I'll believe emacs (*true* emacs: lisp interpreter and everything) when I >see it. Under OS/2, you are *still* limited to 64k segments. > ... >So, tell me: how do you call up your home machine from work to check your >mail? When your housemate is busy running Lotus under DOS, how do you play >with your wonder Logitech debugger? What?! You mean you *can't*?! What an >inferior OS... The fact of the matter is that in the end OS/2 is a single-user, single-machine operating system. There are no real plans afoot at Microsoft to make it otherwise (the claims of the Objects group notwithstanding.) Unix was, from the start, designed to be a multiuser operating system, and it has been evolved into systems which are more-or-less distributed. Unix has become a "portable" system; if one follows a guide like XPG2 then there's reasonable assurance that a program will work on a wide variety of machines. On the other hand, OS/2 was designed specifically around the 80286 memory model (and though OS/2 2.0 will allow for linear address models it still carries around the baggage to handle 286-style applications). *Any* application written for OS/2 is therefore unportable, unless it uses, say, a Posix-compatible library, and then we've essentially guested Unix under OS/2, so what's the point of having OS/2? Another point that isn't often addressed in the OS/2-vs.-Unix debate is system correctness. Ever seen the bug lists for OS/2? Or the QA setup? As much as I rag on SCO all the time I'll take the SCO QA department over the mess at Microsoft. Lastly: OS/2 is a proprietary system. This has lead to some really odd ideosyncrasies in its design (e.g. the event distribution model in PM.) Unix has been reviewed and rewritten and complained about by a lot of people over a number of years, and its design is not so much dependent on the design whims of one or a few people. Here's a challenge for an OS/2 system: build an MIS system with a farm of highly-reliable disks, a pool of processors available for compute- and disk-intensive batch jobs (e.g. end-of-month processing), a set of shared printers with print job routing, and a number of workstations which can access the database on the disk farm. Provide a security system so that access to sensitive data is controlled. For some data, the data must be reasonably secure even during transmission over the network. (This is a thumbnail sketch of a common, *very* simple corporate or local government DP system.) The support for a great many of the tasks involved with building such a system are already provided with a Unix system. You have to do almost all of it yourself under OS/2. Of course you *can* build such a system, but it'll take a lot longer and be an order of magnitude harder to maintain. -richard p.s. Hi Sean! General notes: yes, I used to work for Microsoft. No, I don't like OS/2. As far as my real position on Unix goes I think it's getting creaky with age but I can't think of a better system for immediate commercial use. And yes, I do work on Unix (SunOS, SCO, BSD) now, mostly for X11 work. -- ----------- Richard A. Golding, Crucible (work) and UC Santa Cruz CIS Board (grad student) Internet: golding@cis.ucsc.edu Work: {uunet|ucscc}!cruc!golding Post: Baskin Centre for CE & IS, Appl. Sci. Bldg., UC, Santa Cruz CA 95064