NetNews Usenet Archive 1992 #18/NN_1992_18.iso/spool/comp/os/os2/advocacy/4369

Text File | 1992-08-21 | 9.8 KB | 205 lines

Newsgroups: comp.os.os2.advocacy
Path: sparky!uunet!caen!hellgate.utah.edu!jaguar.cs.utah.edu!brian
From: [e-mail] (Brian Sturgill)
Subject: Re: What and how I'd like to see c.o.o.a advocate.
Date: 21 Aug 92 15:54:55 MDT
Message-ID: <[e-mail]>
Organization: University of Utah CS Dept
References: <[e-mail]> <[e-mail]>
Lines: 194

In article <[e-mail]> [e-mail] (David Charlap) writes:
>In article <[e-mail]> [e-mail] (Brian Sturgill) writes:
>
>>WHERE MICROSOFT WINS
>>Microsoft has the low-end market sewn up... why try to fight it?
>
>There's a wonderful attitude.  With that in mind, we should be still
>using Apple ]['s running VisiCalc.  They had the market sewn up, too.
>Until the PC and Lotus 1-2-3 came along.
But the PC WAS a bigger platform... read below.  The "medium" platform
of the time so to speak.

>
>>Any 386SX machine with small memory, small disk runs Windows just fine,
>>but OS/2's better features make it too fat to run there.
>
>Most Windows apps I've used (like Word, Excel, Harvard Graphics, etc.)
>need a lot more than any 386sx machine with small memory and small
>disk.  Once you expand the computer to handle the software (8MB of
>memory and a big disk), you've got a decent OS/2 platform.  Your
>platform is only adequate for running older application or the
>supplied applets.
Only users that run all this simultaneously hit the 4 megs limit with
these (or people with huge documents or spreadsheets, etc)... these
are precisely the people I'm suggesting OS/2 aim for.
Users of Windows on 4 megs systems that write memos and letters, do simple
drawing and spreadsheets, an occasional print/merge... are quite
well served by Windows.  OS/2 should position itself as the place to
migrate to.  They don't have to wait necessarily for people to use windows
either.  People that use lots of DOS programs and need multitasking are
good targets too.  My argument is that IBM tried to convince people that
OS/2 ran fine in 4 megs... they would be better off to admit the need for
a somewhat higher end system. (And in truth they seem to have started to
do this, but I want to see the official specs changed.)

>
>>NT is clearly designed at the low-end workstation level.  Basically
>>any 386-33Mhz with 12 or 16 megs of memory and higher is prime NT territory,
>>especially if that machine is on a network acting as a server.
>>OS/2 is missing a huge number of features in the area of networking.
>
>No.  You add the LAN server package to get all of those features.
Wrong. Please go back to some of my earlier posts of NT's features...
better yet buy a copy and look for yourself.

>
>>It also does not have security or symmetric multiprocessing.
>
>It has security.  And I know of no PC-compatible machine with hardware
>support for symmetric multiprocessing, so it's not an issue now.  When
>the hardware becomes available, I don't doubt that IBM will have
>updates available for those users who need it.
There are plenty of MIPS-based multiprocessors now available.

... /* Several points of agreement removed... wow, we agree on something! */
>>As things stand developers won't
>>develop the low-end packages, and Microsoft will get around
>>to that mid-range version of Windows and OS/2 will die the death
>>an unresponsive-IBM deserves.
>
>Because of MultiMedia?  It takes more than sound effects to make a
>good OS.  And it takes more than lack of sound effects to break one.
True, but then multi-media is more than sound effects.  Presentation creation
is about to undergo a huge change.  Video and Sound "clip-art" are already
available.  For the class of user under discussion this will be quite
important.

... /* More points of agreement omitted! */
>>Reasonably priced peer-to-peer networking.  Small networks, even
>>at home are fast becoming the norm for an OS/2-class user.
>
>I don't know about the home user.  But small companies definitely
>want these features.  Right now, client-server is the only thing
>provided.
It depends on the "home" user. At the NT developers conference a fair
portion of the people I talked to at lunch were running OS/2 at home
and at the very least had NetWare Lite running between their main machine
and their spouses.  I to was in the same position, my wife and I both
ran OS/2 and used TCP/IP+NFS to share a printer and disk.
Hobbiest are doing this too.  It's just a matter of time till husband and
wife professionals decide they'd like to share stuff.  The big factor is cost.

... /* Yet more agreement omitted! */
>>EITHER DROP OS/2 3.0 OR MAKE IT A SEPARATE PRODUCT
>>It will be impossible to make a symmetric multiprocessing, network
>>server OS that runs well on a current medium-scale machine.  This is not
>>currently a strong point of OS/2.  Why not fix and properly market
>>OS/2 in the market it is sized for?  If IBM wants to compete
>>with NT then do so with AIX... it is currently much better suited to task.
>
>What do you consider medium?  386's?  The multiprocessing and
Medium is what you can buy from Gateway.  Pretty well any 386/486 is
medium.  These machines are NT's low-end.

>portability will be based on the Mach MicroKernel (the last I heard).
>This already runs on the 386 platform.  Why shouldn't IBM exploit that
>end?  AIX will also run on top of Mach at that point.  Why not
>continue to sell both products for both platforms (AIX/386 exists).
The Mach microkernel is a pitiful place to put OS/2... it is slow in
design.
Mach chose a model whereby each system call forces 2 context switches, and
on any currently known/available architecture, context switches are expensive.
This is particularly true on RISC systems.
There are other elements in the Mach's design that falls along the same lines.
Mach was designed to meet a religious idea of what a MicroKernel is.

For example consider under OSF/2 a read system call:
    {User Process}   {File Server Process}  {Security Server Process}
-----------------border between kernel and user execution rings----------
            {Micro Kernel}

The User Process calls through the MK to the File Server (a context switch),
The File Server calls the Security Server to authenticate the user
(a context switch), which does the authentication and returns to the File Server
(a context switch), which schedules the read, and eventually returns
to the User Process (a context switch).

This flaw exists even in the simpler Mach that OSF/1 is based on.  Thus
IBM has not replaced AIX with OSF/1 (which is what they had intended earlier).
Note... the Mach in NextStep, is an early version of Mach that did not
really have a microkernel, but rather had a micro-kernel design.  It's
implementation is as a monolithic kernel, and thus does not have this
problem.

FYI, NT looks like this:

    {User Process}
-----------------border between kernel and user execution rings----------
    {File Server, MicroKernel, Security Server}

The same read call will trap to the kernel ring and then execute with
the approximate overhead of function calls between "servers".
There are no context switches.
A couple of European-based "micro-kernel" designs share a similar flavor...
I forget their names.

>
>Also note that OS/2 3.0 is a "future goal".  There is no release date
>or anything else about it other than some nice ideas.  It's kind of
>hard for a company to drop a product that doesn't even exist.
I am confused about why they would want to create an OS/2 that's at
the same level as AIX... these products would clearly compete.
I would rather them not waste time on making the ultimate server OS/2, but
rather create the ultimate desktop platform.

>
>>As currently designed, OS/2 is much more suited to being a personal
>>operating system.  Why not capitalize on that as a feature.
Where are they capitalizing on it... what are you seeing I don't?
Where is the advertising that says it's the "DOS" of your dreams.
Simultaneously you can run 5 different DOS programs, etc.  The most
infomative thing I've seen in the advertising is "It's not about
closing old windows, but opening new doors".  Rather vague.
I guess I saw one extended ad in Corporate Computing... it was too long
so I didn't read it... I did however read the editorial in the back
cricizing OS/2.

>
>They are.  But you just criticized it for not having networking.  What
>do you want them to do?
Networking is not the same as being a server OS.  I mean they should
concentrate on it being a good client to networks.  I also said it
would be ok to charge for the networking piece, just that it should
be cheap to do simple peer-to-peer.

>
>>FINALLY
>>IBM should make a multi-tiered plan that says what's for the low, medium,
>>and high-end, and quit confusing everybody about whether and how Taligent,
>>OS/2, and AIX fit together.
>
>They have done that for a long time:
>
>PC-DOS for the bottom-end PC's.
>OS/2 for the 386 based systems and better Intel platforms
>AIX for the workstations (read: RS-6000)
>OS/400 for their AS/400 minis
>MVS or VM for their ES-9000 mainframes
This was the plan some time ago.  But isn't Taligent going to be the Ultimate
desktop OS?  How does a portable, symetric multiprocessing, secure version
of OS/2 2.0 differ from AIX?  To me the waters look very muddy.

>
>All can talk to each other very nicely using SNA networking, which all
>IBM operating systems can make use of.  Taligent isn't fitting into
>anyone's plan for today - it's another future product.
Fine, if you're a Big Blue site.  More and more, most places aren't.
Ah, but if I buy into OS/2, what will happen when Taligent comes out?
They've already stated via press release that OS/2 2.0 support will not
be in the initial release of Taligent, but will follow in 1995.
Taligent is scheduled for 1Q 1994, but is ahead of schedule and looks
like it will be out 4Q 1993... that's a hell of a long gap for 2.0 users.

Brian